Other Computer Science · Diting Lab

Average pairwise Hamming distance within hypercube subsets

2026-10-30T00:00:00+09:00

Average pairwise Hamming distance of a subset in a hypercube arXiv · Oct 30, 2026This paper studies subsets of the hypercube {0,1}^n with density 1/8 and proves that their average pairwise Hamming distance is at least (n-3)/2. Equality holds only for (n-3)-dimensional subcubes. The result answers a question posed by Lei Yu and resolves an important open case of a broader problem raised by Ahlswede and Katona in 1977. ★★★★ Notable: The authors prove that a density-1/8 subset of the hypercube has average pairwise Hamming distance at least (n-3)/2, with equality only for (n-3)-dimensional subcubes, answering a question of Yu and a

Finding equilibrium prices in online markets with fees, taxes, and subsidies

2026-10-30T00:00:00+09:00

Finding Competitive Equilibria in Online Marketplaces with Platform Fees, Taxes, and Subsidies arXiv · Oct 30, 2026The paper studies how to compute market equilibria in multi-unit trading networks where valuations satisfy generalized full substitutes and transfers are piecewise-linear and specific to agents and transactions, modeling platform fees, taxes, subsidies, and membership discounts. The authors propose an ascending price procedure that raises prices at different rates and reaches the minimum equilibrium price vector in finitely many iterations, each implementable in strongly polynomial time. An equilibrium allocation can then be recovered efficiently. The approach extends primal-dual auction techniques to handle complementarities between purchases and sales via a submodular flow formulation. ★★★★ Notable: The authors extend primal-dual ascending auction techniques to trading networks with platform fees and subsidies, reporting strongly polynomial equilibrium computation.

Nearly optimal sample-size versus consensus-time trade-off in consensus dynamics

2026-10-30T00:00:00+09:00

Nearly-Optimal Time-Sample Trade-off in Consensus Dynamics arXiv · Oct 30, 2026This paper studies how the number of sampled opinions affects the time needed for distributed nodes to reach consensus. It analyzes the synchronous h-Majority rule, in which each node samples h random nodes per round and adopts the most frequent opinion, with random tie-breaking. The authors show that h-Majority achieves a nearly optimal trade-off between sample size and consensus time among a broad class of consensus dynamics. ★★★★ Notable: The authors give a consensus-time upper bound for h-Majority valid for all h≥3 and initial configurations, with a matching lower bound up to polylog factors.

A sharp Fisher-information inequality for Boolean pairs, proving the Amari–Kobayashi conjecture

2026-10-29T00:00:00+09:00

A Sharp Fisher-Information Inequality for Boolean Pairs: Proof of the Amari--Kobayashi Conjecture arXiv · Oct 29, 2026The paper considers two terminals that observe correlated uniform binary vectors and each compress their observation to one bit. It proves that, for every blocklength and correlation parameter in (-1,1), the Fisher information kept by the two bits is maximized by matching parity functions of a suitable degree, resolving a conjecture of Amari and Kobayashi. The proof combines a Fourier-based bound on the derivative of the normalized covariance with a scalar inequality for signed polynomials under a global range constraint. ★★★★ Notable: The authors prove the Amari–Kobayashi conjecture on Fisher information of Boolean pairs, combining a Fourier bound with a signed-polynomial inequality for all blocklengths.

Proximity Gaps for One-Point Algebraic Geometric Codes

2026-10-29T00:00:00+09:00

On Proximity Gaps for One-Point Algebraic Geometric Codes arXiv · Oct 29, 2026This paper studies proximity gaps and correlated agreement for general one-point algebraic geometry (AG) codes, properties that underpin the soundness of interactive oracle proofs. It extends known results for Reed-Solomon codes, proving correlated agreement and soundness bounds for these codes. Using Garcia-Stichtenoth towers, it builds asymptotically good code families and gives two counterexamples showing limits beyond half the minimum distance. ★★★★ Notable: The authors extend proximity gap and correlated agreement results from Reed-Solomon to general one-point AG codes, with length-independent soundness and counterexamples.

Tight Bounds for Anonymous Parity Certification

2026-10-29T00:00:00+09:00

Tight Bounds for Anonymous Parity Certification arXiv · Oct 29, 2026This paper studies how many bits per vertex are needed to certify that a graph has even order in the anonymous radius-one local certification model. The authors prove the complexity on general graphs is Θ(log log n) bits, closing a gap left by prior work. They also show graphs of degeneracy at most d can be certified with O(log d) bits, and identify nowhere-dense classes where constant-size certificates cannot be guaranteed. ★★★★ Notable: The authors close the gap to a tight Θ(log log n) bound for anonymous radius-one even-order certification and characterize degeneracy and nowhere-dense classes.

Sensor-side hardwired units for parallel vision pre-processing that frees the host CPU

2026-10-29T00:00:00+09:00

Hardwired Distributed Processing Units at the Sensor: Truly Parallel Vision Pre-Processing that Frees the Host CPU arXiv · Oct 29, 2026The paper proposes placing small, fixed hardware datapaths (HDPUs) directly at the image sensor, each performing one always-on task in parallel on the pixel stream and sending only compact results to the host. The authors build three units (Canny edges, Harris corners, and a line tracker that issues steering decisions) on a low-cost Artix-7 FPGA using 2,568 LUTs and an estimated 0.318 W. A full-system simulation compares these units with a frame-based host program, showing much earlier responses and lower host CPU load. ★★★★ Notable: The authors propose sensor-attached hardwired units for parallel vision pre-processing, implement them on an Artix-7 FPGA, and report full-system simulation comparisons against host latency.

Upper bounds on CSP non-redundancy using locally decodable codes

2026-10-29T00:00:00+09:00

Upper Bounds for CSP Non-redundancy via Locally Decodable Codes arXiv · Oct 29, 2026This paper studies non-redundancy in constraint satisfaction problems (CSPs), which measures how many logically independent clauses a CSP instance can contain. The authors present a new method that uses locally decodable codes (LDCs) and locally correctable codes (LCCs) to upper-bound the non-redundancy of certain relations. Applying this approach, they obtain improved upper bounds for the BCK predicate and its Boolean variant, and give the first Boolean predicate whose non-redundancy provably is not of the form n^{k±o(1)}. ★★★★ Notable: The authors bound CSP non-redundancy via 2-query LDC and LCC structure, improving BCK bounds and reporting a first Boolean predicate outside the n^{k±o(1)} form.

SCL-Aware Design of PAC Codes: Designing Error-Correcting Codes for List Decoding

2026-10-30T00:00:00+09:00

An SCL-Aware Design of PAC Codes arXiv · Oct 30, 2026This paper presents design methods for the rate profile of polarization-adjusted convolutional (PAC) codes intended for successive cancellation list (SCL) decoding. It introduces a conditional density evolution Gaussian approximation (CDEGA) to analyze decoding with errors, and a dangerous-path occupancy metric to measure bit-error occupancy in a finite list. Using these tools, the authors design codes of length up to 512 that reportedly achieve at least half a dB gain over traditional PAC codes at an FER of 10^-3.

SPECTRUM: A Blueprint and Strategy for Europe's Exascale Research Data and Compute Infrastructure

2026-10-30T00:00:00+09:00

SPECTRUM: A Strategic Framework and Technical Blueprint for European Exascale Research Data and Compute Infrastructure arXiv · Oct 30, 2026This paper summarises two SPECTRUM project documents: a Technical Blueprint and a Strategic Research, Innovation and Deployment Agenda (SRIDA). It notes that the HiLumi LHC, SKA Observatory and LOFAR 2.0 will push high energy physics and radio astronomy towards exabyte-scale data within a decade. The documents share the same evidence base, offering technical recommendations and a phased roadmap of priorities. The paper also describes their consultative method and the impact the SRIDA expects.

Can Language Models Be Trusted to Orchestrate Edge-Cloud Services?

2026-10-30T00:00:00+09:00

Can We Rely on Language Models to Orchestrate Services across the Edge-Cloud Continuum? arXiv · Oct 30, 2026This paper systematically reviews 39 primary studies on using language models to orchestrate services across the edge-cloud continuum. It proposes a five-dimensional framework linking system context, orchestration decisions, language model agency, operational safeguards, and evaluation evidence. The review finds mostly hybrid designs, where language models support interpretation, candidate generation, and mitigation planning, while downstream components typically keep feasibility enforcement and execution authority.

First-order bilevel optimization for designing non-truthful auctions

2026-10-30T00:00:00+09:00

First-Order Bilevel Optimization for Non-truthful Auction Design arXiv · Oct 30, 2026The paper formulates auction design as a Stackelberg game with multiple strategic followers and proposes a fully first-order algorithm for revenue optimization in non-truthful auctions. It establishes non-asymptotic guarantees for hypergradient approximation and shows that, under standard smoothness and strong-monotonicity assumptions, the method reaches an O(ε)-stationary point with about Õ(ε⁻²) first-order oracle calls. Experiments on single-item and multi-item auctions report faster convergence than second- and zeroth-order baselines, better scaling with more bidders, and auctions whose revenue approaches that of truthful baselines.

True 3D Holography: Future Communication Service and Requirements for Converged Cloud-Network Architecture toward 6G

2026-10-30T00:00:00+09:00

True 3D Holography: A Communication Service of Tomorrow and Its Requirements for a New Converged Cloud and Network Architecture on the Path to 6G arXiv · Oct 30, 2026This paper examines true 3D holographic communication as a use case in the 6G NeXt research project and sets out its requirements for communication and computing infrastructure. It argues that such services need high processing power, a high-speed distributed backbone supporting split computing, and tight integration between processing facilities and wireless networks. The authors propose an overall communication and computing architecture and discuss novel technological approaches based on it.

Measuring How Much Communication Helps Multi-Agent Target Tracking at Limited Data Rates

2026-10-29T00:00:00+09:00

Measuring the Value of Communication in Multi-Agent Tracking Under Finite-Rate Belief Sharing arXiv · Oct 29, 2026This paper studies how useful communication is for stationary agents cooperatively tracking a moving target when they can only exchange information at limited rates. Each agent uses range and bearing measurements with an extended Kalman filter, quantizes its state estimate, and fuses shared information via covariance intersection. Results show very low rates can hurt performance due to coarse quantization, while moderate rates capture most of the gain, with diminishing returns at higher rates.

PhOSv1: Scaling model sleeping on one server by sharing GPU contexts for faster autoscaling

2026-10-29T00:00:00+09:00

PhOSv1: Efficient Scaling of Model Sleeping on a Single Server for Autoscaling with GPU-Context Deduplication arXiv · Oct 29, 2026This paper addresses autoscaling in model-as-a-service systems, where existing model sleeping methods keep each sleeping model's GPU context resident, so the cost grows with the number of models. The authors present PhOSv1, which routes each engine's GPU context allocation through a central daemon via API remoting, deduplicating shared contexts. This lets a single server keep many more models asleep. The evaluation reports scale-up time matching vanilla vLLM sleep mode while substantially reducing GPU memory footprint.

An exact parallel dense Gauss-Seidel method trading parallel runtime for preprocessing time

2026-10-29T00:00:00+09:00

An Exact Parallel Dense Gauss-Seidel Algorithm with Trade-Offs between Parallel Execution Time and Preprocessing Time arXiv · Oct 29, 2026The Gauss-Seidel method is widely used for solving linear equation systems because it converges well, but its data dependencies make parallelization difficult. This paper proposes a new parallel Gauss-Seidel algorithm that achieves better parallel runtimes in exchange for more preprocessing time. The algorithm can be tuned to offer trade-offs between preprocessing time and the best achievable parallel runtime.

TRACEs: An Embedded Provenance Architecture for Distributed Scientific Workflows

2026-10-29T00:00:00+09:00

TRACEs: Fine-Grained Embedded Provenance for Distributed Scientific Workflows arXiv · Oct 29, 2026Large-scale scientific workflows run on distributed HPC resources, which makes it hard to explain how data products are generated and why similar runs give different results. TRACEs is a provenance architecture for distributed containerized workflows that embeds data lineage and execution context directly in every generated data artifact, and it stays compatible with existing workflow management systems. Evaluated on a seven-stage E. coli variant-calling workflow across multiple HPC nodes, it achieved 100% provenance completeness under distributed execution. It also traced a 200-fold difference in variant counts to differences in input sequencing data without re-executing the workflow.

Level crossing rate bounds for products of independent nonstationary fading processes

2026-10-29T00:00:00+09:00

Level Crossing Rate Bounds for Products of Independent Nonstationary Fading Processes arXiv · Oct 29, 2026This paper develops a unified framework for the instantaneous total level crossing rate (LCR) of products of independent positive nonstationary fading processes. Universal upper and lower bounds are derived from an exact decomposition of the product derivative and signed crossing imbalance, using only factor-wise crossing statistics. The authors introduce a drifting Gaussian-derivative (DGD) class, giving bounds via an exact Gaussian-mixture representation that capture aggregate drift and derivative fluctuations. Numerical results show that the DGD bounds closely bracket the exact LCR.

Optimal self-orthogonal binary linear codes

2026-10-29T00:00:00+09:00

Optimal self-orthogonal binary linear codes arXiv · Oct 29, 2026The paper completes the determination of the minimum possible length of eight-dimensional binary self-orthogonal codes with a given minimum distance. For nine-dimensional codes it gives exact values and bounds for minimum distance up to 64. Using a geometric view of linear codes as multisets of points, the authors redo the classification for dimensions up to seven without computer assistance. They also prove that [93,9,44]_2 codes do not exist.

Information geometry of Sinkhorn divergence on a Hilbert manifold

2026-10-29T00:00:00+09:00

The information geometry of the Sinkhorn divergence on a Hilbert manifold arXiv · Oct 29, 2026This paper studies the metric tensor that the Sinkhorn divergence, a debiased entropic optimal transport cost, induces on probability measures defined along curves. The authors construct a real-analytic Hilbert manifold on which this tensor is a weak Riemannian metric, and extend Eguchi's geometric construction to third order. They also show the metric is bounded above by a multiple of the Fisher-Rao metric on the smooth stratum.

Dynamic Resource Scaling for Quantum Circuit Simulations on HPC Systems

2026-10-29T00:00:00+09:00

Enabling Dynamic Resource Scaling in Quantum Circuit Simulations arXiv · Oct 29, 2026This paper makes the QuEST state-vector simulator malleable using the DMR framework, so one job can resize its node allocation between quantum circuits with different qubit counts. Because no state vector persists between circuits, resizing needs no data redistribution, and a checkpoint of the next circuit's index allows resumption. The authors evaluate reconfiguration overhead on MareNostrum 5 and show that a single job can run circuits of varying sizes with negligible overhead.

Support-distribution conditions for MSRD codes reduce to one inequality

2026-10-29T00:00:00+09:00

The Support-Distribution Conditions for MSRD Codes Reduce to a Single Inequality arXiv · Oct 29, 2026This paper studies linear maximum sum-rank distance (MSRD) codes and counts codewords with a prescribed nonzero support. It proves that a family of nonnegativity conditions, conjectured by Byrne, Gluesing-Luerssen, and Ravagnani, is equivalent to a single condition at the vector of total dimension d+1 obtained by filling the longest blocks first. The result yields an explicit existence criterion over every finite field. The paper also shows an infinite family of parameters that satisfies the joint condition yet admits no such linear code.

Energy-efficient uplink RSMA using a Rydberg atomic quantum receiver

2026-10-29T00:00:00+09:00

Improving Energy Efficiency of Uplink RSMA via Rydberg Atomic Quantum Receiver arXiv · Oct 29, 2026This paper studies an uplink rate-splitting multiple access (RSMA) network that uses a Rydberg atomic quantum receiver (RAQR) to maximize energy efficiency. To handle the direct-readout RAQR's amplitude-only measurements, the authors propose a phase alignment mechanism that linearizes the received signal while keeping channel energy. For symmetric rate requirements, they derive closed-form power allocation from KKT conditions, while for asymmetric requirements they propose a block coordinate descent power allocation and a low-complexity paired allocation scheme. Simulations show the KKT-based scheme achieves the highest energy efficiency, and the low-complexity scheme cuts computation with moderate efficiency loss.

Length limits for MDS algebraic geometry codes from curves of any positive genus

2026-10-29T00:00:00+09:00

Length Bounds for MDS Algebraic Geometry Codes of Arbitrary Positive Genus arXiv · Oct 29, 2026This paper studies the length of maximum distance separable (MDS) algebraic geometry codes built from curves of arbitrary positive genus. For fixed genus, it proves that the ratio of length to field size is asymptotically at most 1/2 when both dimension and codimension grow faster than the logarithm of the field size. It also gives explicit bounds under logarithmic dimension conditions. The bounds hold in every characteristic and do not depend on the choice of curve or evaluation set. The proof combines character sums on the Jacobian with an estimate for sums of distinct points.

A uniform upper bound on the length of MDS codes from genus-two curves

2026-10-29T00:00:00+09:00

A Uniform Upper Bound on the Length of MDS Codes from Genus-Two Curves arXiv · Oct 29, 2026This paper studies the maximal length of MDS algebraic geometry codes built from genus-two curves over finite fields. For codes of length n and dimension k with 3≤k≤n-3, the authors prove n<(2q+10√q+16)/3 for every prime power q, regardless of the parity of k. This gives a uniform upper bound of 2q/3+O(√q), which is strictly smaller than q+1 when q≥125.

Secure wireless communication using flexible reflectors: joint optimization of their position and rotation for physical-layer security

2026-10-29T00:00:00+09:00

Secure Wireless Communications with Flexible Reflectors: Joint Position and Rotation Optimization for Physical Layer Security arXiv · Oct 29, 2026This paper studies secure communication over non-line-of-sight links aided by flexible reflectors (FRs), passive metallic plates on movable platforms, and jointly optimizes their positions and rotations to maximize the secrecy rate. For a single FR, it proposes a Fibonacci segmentation search (FSS) method and a lower-complexity cross-entropy search (CES) method, then extends both to multiple FRs. Simulations show FSS achieves significant secrecy rate gains over benchmarks, while CES offers a favorable runtime-performance tradeoff.

Speeding up finite Markov chains via Bregman projections and partition design

2026-10-29T00:00:00+09:00

Bregman projections and partition design for accelerating finite Markov chains arXiv · Oct 29, 2026The paper studies Gibbs averaging of finite Markov chains with a prescribed stationary distribution, built from a Gibbs kernel induced by a state-space partition. It proves that the projected chain GPG is an information projection for admissible Bregman generators, with a Pythagorean identity and a dual characterization. For reversible ergodic baselines, Gibbs averaging does not increase the squared weighted Frobenius distance to stationarity, so the mixing time does not increase. The paper also gives partition-selection objectives and algorithms, and a numerical experiment compares the projected chains with a local Metropolis sampler.

Fragmented Obedient Equilibria in Routing Games: a new equilibrium concept for competing mediators recommending routes

2026-10-29T00:00:00+09:00

Fragmented Obedient Equilibria in Routing Games arXiv · Oct 29, 2026The paper studies routing games where competing mediators recommend routes to fixed nonatomic populations. Each mediator chooses a randomized flow allocation to minimize its population's expected travel cost, subject to users having no conditional expected gain above a tolerance ε from ignoring recommendations. The authors define fragmented obedient equilibria (FOE) and prove existence under a lower-hemicontinuity assumption on obedient policy correspondences. Under affine edge latencies, the problem reduces exactly to first and second flow moments, and for networks with at most four routes each best response is a semidefinite program. A Braess network example shows that relaxing obedience can lower FOE costs, though gaps to the social optimum and optimal obedient pooling can persist.

When Heterogeneity Helps: GPU and WSE Complementarity in Systems

2026-10-29T00:00:00+09:00

When Does Heterogeneity Pay Off? Understanding GPU-WSE System Complementarity arXiv · Oct 29, 2026This paper studies unified systems that combine GPUs and wafer-scale engines (WSEs) for AI workloads. The two differ in compute, memory, and communication organization, which creates both architectural complementarity and extra overheads. The authors propose a hierarchical benchmarking methodology to characterize GPU-WSE heterogeneous execution and the conditions for exploiting complementarity. They also present an analytical performance model that identifies when heterogeneity becomes a bottleneck and when it yields system-level gains.

PDScope: Resolving the Prefill-Decode Trade-off in Memory-Limited MoE Inference

2026-10-29T00:00:00+09:00

PDScope: Breaking the Prefill-Decode Seesaw for Memory-Bound MoE Inference arXiv · Oct 29, 2026This paper addresses local deployment of MoE-based LLMs on CPU-GPU nodes, where limited GPU memory forces most experts into CPU memory. Prefill is compute-bound and favors aggressive GPU loading, while decode is I/O-bound and needs conservative transfers, creating a trade-off under low batching. PDScope decouples the two phases, predicting expert activations and optimizing placement, prefetching, and execution for each, using a static computation graph, cache prefetching, and zero-copy asynchronous I/O. On mainstream MoE models, it reports 2.5-10.3x higher prefill throughput and 2.1-6.1x higher decode throughput than existing systems.

Phase retrieval of bandlimited signals via Morlet and companion wavelets

2026-10-29T00:00:00+09:00

Phase Retrieval of Bandlimited Signals Using a Morlet Wavelet and an Admissible Companion arXiv · Oct 29, 2026The paper shows that complex bandlimited signals can be uniquely recovered, up to a global phase factor, from two families of wavelet transform magnitudes. These families arise from translations of a fixed dilated Morlet wavelet and a fixed companion wavelet, sampled at twice the Nyquist rate in two frequency bins. The authors reformulate both families as modulated Gaussian convolution measurements of an auxiliary bandlimited signal and apply a Gaussian convolution uniqueness result. They prove that samples on (4B)^{-1}Z determine every signal in PW_B^2, using significantly fewer sampling points than earlier wavelet phase retrieval methods.

Consumer welfare limits of market segmentation when some consumers are captive

2026-10-29T00:00:00+09:00

The Welfare Limits of Market Segmentation with Captive Consumers arXiv · Oct 29, 2026The paper studies a duopoly with captive and contested consumers and characterizes every pair of consumer and producer surplus achievable under some market composition and some public segmentation, constructing a market for each pair. The achievable set is lens-shaped. At any industry profit level, consumer welfare is lowest when captive and contested consumers are perfectly separated, and highest in an unsegmented market where the two firms have equal numbers of captives.

Trade-offs between communication and privacy in distributed storage using shared secret keys

2026-10-29T00:00:00+09:00

Communication and Privacy Trade-Offs in Distributed Storage With Shared Secret Keys arXiv · Oct 29, 2026This paper studies distributed storage in which all transmissions occur over a public channel, with privacy provided by shared secret keys. A user stores a file across L servers so that any t servers can reconstruct it, while any z colluding servers learn at most a fraction α of the file. For a fixed key length n, the authors characterize the trade-off between the privacy leakage α and the maximum securely storable file size. They also determine the minimum local randomness, public communication, and storage needed to achieve this optimal trade-off.

Radar sensing of fast, distant targets in overspread channels using Zak-OTFS and modular arithmetic

2026-10-29T00:00:00+09:00

Sensing Fast and Distant Targets in Overspread Channels with Zak-OTFS and Modular Arithmetic arXiv · Oct 29, 2026This paper addresses aliasing in delay-Doppler (DD) modulation when the channel spread factor exceeds one, which is common in long-range mobility scenarios such as drones and satellites. The authors propose a two-frame design that combines DD bases with different delay and Doppler periods on the same bandwidth and frame size, and use modular arithmetic on wrapped peak positions to recover each target. They introduce two receivers, DS and DS-XC, with DS-XC including cross-pilot cancellation. In a 100 km drone scenario, DS-XC detects 98% of targets, compared with 81% for the best baseline.

Generative rate-distortion-perception theory on Riemannian manifolds: how much information is needed for generated outputs to stay accurate on curved spaces.

2026-10-29T00:00:00+09:00

Generative Rate-Distortion-Perception Theory on Riemannian Manifolds arXiv · Oct 29, 2026The paper studies how much source information must be sent so that a receiver's generated outputs meet fidelity and reliability requirements on compact Riemannian manifolds, while preserving the source distribution. It proposes a generative rate-distortion-perception framework without common randomness and gives a single-letter characterization of the asymptotic operational function. For uniform sources on compact homogeneous manifolds, geometric lower and upper bounds coincide, giving an exact limiting result. On unit spheres, it shows that for dimension d≥2, adding d bits per source symbol can reduce the distortion radius by more than half.

DistOptimize: Automatic Optimizer for Distributed Protocols

2026-10-29T00:00:00+09:00

DistOptimize: Automatic Optimization of Distributed Protocols arXiv · Oct 29, 2026This paper presents DistOptimize, an automatic optimizer for distributed protocols that avoids safety risks from manual changes. It adapts throughput optimizations from academic languages to production frameworks, verifies the rewrites' correctness in Rocq, and automates their application using calibration and ILP-based search. Applied to fundamental protocols and protocols used at AWS, it improves throughput by up to 4x while increasing latency by at most 0.3ms.

RARE: A new equilibrium for pricing games where sellers can respond to price changes

2026-10-29T00:00:00+09:00

RARE: Response-Aware Rational Equilibrium in Pricing Games arXiv · Oct 29, 2026This paper studies pricing games in markets where budget-constrained buyers choose utility-maximizing bundles and each seller sets a price for one good. It proposes Response-Aware Rational Equilibrium (RARE), in which any seller's deviation can be deterred by other sellers' responses, and shows that RARE exists under certain conditions, with approximate versions for some cases. The authors also analyze a price-raising dynamic that reaches an approximate RARE.

Principal-agent hypothesis testing with adaptive, sequential experimentation

2026-10-29T00:00:00+09:00

Principal-agent hypothesis testing under adaptive experimentation arXiv · Oct 29, 2026This paper studies principal-agent hypothesis testing, where a principal rewards an agent based on data the agent generates. It extends the single-shot framework to sequential settings, where evidence accumulates over time, rewards may be deferred or paid incrementally, and the agent can stop the contract at a strategically chosen time. The main result shows that such contracts, called null-deterring, are characterized by e-processes.

Random Non-Commutative Error-Correcting Codes: Extending Random Coding Theory to Arbitrary Groups

2026-10-29T00:00:00+09:00

Random Non-Commutative Error-Correcting Codes arXiv · Oct 29, 2026This paper builds on the non-commutative error-correcting codes introduced by Chapman, Dinur, and Lubotzky, studying random codes over arbitrary groups as an analogue of classical random coding. It examines how far the classical Gilbert-Varshamov (GV) bound extends to this setting, showing that random universal codes of linear size in free groups are asymptotically bad. The authors characterize exactly which families of finite groups admit good random codes, namely those with uniform polynomial maximal-subgroup growth. They also give explicit quantitative bounds on rate and optimal distance based on intrinsic group parameters, and prove that families of finite solvable groups always form GV families.

Slant: Deliberately Uneven Traffic Routing to Save Resources in Microservices

2026-10-29T00:00:00+09:00

Slant: Shaping Traffic for Resource-Efficient Microservices arXiv · Oct 29, 2026Load balancers in microservice systems usually aim for even request distribution without considering how routing affects autoscaling signals and resource efficiency. This paper presents Slant, a resource-aware load balancing framework that intentionally creates safe traffic imbalance to improve utilization and works with existing vertical pod autoscalers without modification. In experiments on three microservice benchmarks, Slant reduces allocated CPU cores by up to 26% and lowers P99 latency by up to 18% compared to Round-Robin.

A paper on a unified system for managing data, metadata and paradata of 3D digital cultural heritage objects, covering automated workflows, federated login and publishing.

2026-10-28T00:00:00+09:00

Integrated data, metadata, and paradata management system for 3D Digital Cultural Heritage objects: Workflow automation, federated authentication, and publication arXiv · Oct 28, 2026The paper addresses the difficulty of managing complex, high-quality 3D digitised cultural heritage objects and their interconnected metadata. It proposes a global federated authentication and authorisation mechanism, a data and metadata management system, and an integrated engine for designing and running automated workflows. The solution is evaluated on three distinct 3D digitised cultural objects, showing the full process from data upload to publication.

Dimension-free exponential entropy quotients over the field F₂

2026-10-28T00:00:00+09:00

Dimension-free exponential entropy quotients over 𝔽₂ arXiv · Oct 28, 2026This paper studies entropy quotients for random vectors over F₂. For every ε>0, the authors show there is a subspace G with dimension at most 2^{320(1+ε)/ε} times H(X) such that H(q_GX) is at most (1+ε)(H(q_GX+q_GX')−H(q_GX)). The coefficient does not depend on the ambient dimension, and its exponential order in 1/ε is shown to be necessary.

Determining the weight enumerators of binary extremal self-dual codes of length 60

2026-10-28T00:00:00+09:00

Determination of weight enumerators of binary extremal self-dual codes of length 60 arXiv · Oct 28, 2026The paper constructs three binary extremal self-dual codes of length 60 whose weight enumerators were not previously known to occur. This completes the determination of all weight enumerators that arise for binary extremal self-dual codes of length 60.

Random Centres and the ε-Dependence of Constant-Round Planar Domination

2026-10-28T00:00:00+09:00

Random Centres and the ǎrepsilon-Dependence of Constant-Round Planar Domination arXiv · Oct 28, 2026This paper studies constant-round LOCAL algorithms for minimum dominating set on planar graphs. It replaces the heaviest-edge star selection in the clustering step with a random-centre rule, which reduces the round-complexity exponent. The authors also analyse how the approximation parameter ε affects the number of rounds, showing Ω(1/ε) rounds are needed on paths and Θ(1/ε) rounds are optimal on forests and outerplanar graphs. ★★★★ Notable: Replaces heaviest-edge star selection with random centres, reducing the planar round exponent from about 25.81 to below 5.89, with lower bounds reported.

Simultaneous Gradient Learning in First-Price Auctions: How Gradient-Based Learning Converges to Equilibrium Under Private Information

2026-10-28T00:00:00+09:00

Simultaneous Gradient Learning in First-Price Auctions arXiv · Oct 28, 2026The paper uses first-price auctions as a testbed to study equilibrium learning in discretized Bayesian games. It introduces Bayesian semicoarse correlated equilibrium (BSCCE), a refinement of Bayesian coarse correlated equilibrium (BCCE), and gives a linear programming characterization. The main finding is that the BSCCE set contracts in many cases as the discretization is refined, with its polytope diameter shrinking toward zero and converging to the unique continuous-game Bayes-Nash equilibrium.

Hybrid multiple access for uplink communication using atomic quantum receivers to reduce transmit power

2026-10-28T00:00:00+09:00

Rydberg Atomic Quantum Receiver Empowered Hybrid Non-orthogonal Multiple Access arXiv · Oct 28, 2026This paper proposes an uplink hybrid NOMA architecture assisted by Rydberg atomic quantum receivers to improve energy efficiency. It formulates a joint optimization of phase measurements and multi-user power allocation to minimize total transmit power, solved with a three-layer hierarchical decomposition framework. Simulations report about a 93% reduction in total power consumption compared with OMA baselines while maintaining backward compatibility.

A π/4 lower bound on recovery time for random access in symmetric linear codes

2026-10-28T00:00:00+09:00

A π / 4 lower bound for random access in symmetric linear ensembles under optimal decoding arXiv · Oct 28, 2026The paper studies how many random linear observations are expected before a specific information symbol can be recovered. For ensembles with uniformly random supports, independent nonzero coefficients, and bounded column weights, it proves that the asymptotic normalized expected recovery time is at least π/4 under optimal decoding. The bound holds over finite or infinite fields and extends an earlier result for binary fully symmetric codes under peeling decoding.

MacKay-Neal Codes for DNA Channels Under Block-MAP Decoding

2026-10-28T00:00:00+09:00

MacKay-Neal Codes for DNA Channels under Block-MAP Decoding arXiv · Oct 28, 2026This paper studies MacKay-Neal socket codes for a quaternary insertion-deletion-substitution channel that models DNA reads, decoded with block maximum a posteriori (MAP) decoding. It proves that for a family of fixed-degree socket ensembles, the limiting rate is 6/ℓ bits per nucleotide, and that the ensemble-average block error is O(1/n) when the joint symmetric information rate of a fixed number of reads exceeds this rate. Deterministic graph-offset sequences are shown to achieve the same limiting rate and error bound.

Polynomial-time constructions of asymptotically good random access codes

2026-10-28T00:00:00+09:00

Polynomial-Time Constructions of Asymptotically Good Random Access Codes from Linear Codes arXiv · Oct 28, 2026Random access codes compress a q-ary input string into fewer q-ary symbols so that any requested input symbol can be recovered with probability above the random-guessing level 1/q. The paper introduces asymptotically good RACs and gives two polynomial-time constructions: OA-RACs built from linear codes via their orthogonal-array property, and a direct-sum strategy using q-ary Hamming codes that yields asymptotically good families at every rate in (0,1). Numerical results show favorable finite-length performance for both constructions.

Is dual polarization always optimal? A study of new spatial degrees of freedom

2026-10-28T00:00:00+09:00

Is Dual Polarization Always Optimal? Unlocking New Spatial Degrees of Freedom arXiv · Oct 28, 2026This paper presents a modular, low-cost tripolarized antenna operating in the 2.4 GHz band to experimentally test theoretical findings about tripolarized antennas. Experiments were conducted in two indoor scenarios and one outdoor scenario, with a multi-antenna transmitter and a receiver using a single tripolarized antenna. The authors report that this antenna can increase the number of spatial streams up to three or extend the region where terminals achieve two spatial streams.

FlatCake: A flat storage design for reusing LLM attention caches across host memory and SSD

2026-10-28T00:00:00+09:00

FlatCake: A Flat KVCache Storage Architecture for LLM Serving arXiv · Oct 28, 2026FlatCake is a storage architecture for LLM serving that treats host DRAM and NVMe SSD as equal peer media in one unified address space, instead of a tiered cache. Each KV block is placed at creation directly on its target medium in session-contiguous, layer-major order and is never moved. On a production Claude Code trace, the authors report a 94% cache hit rate under increasing load (versus 64% for Mooncake) and a 1.6× reduction in peak-load mean TTFT.

Exact second-order asymptotics for the Gaussian Gray-Wyner problem

2026-10-28T00:00:00+09:00

Exact Second-Order Asymptotics for the Gaussian Gray--Wyner Problem arXiv · Oct 28, 2026This paper revisits the Gray-Wyner problem of lossy compression for bivariate Gaussian sources under quadratic distortion and derives exact second-order asymptotics. It first identifies boundary rate tuples using the explicit first-order results of Yu (TIT 2023), then characterizes the tradeoff among three second-order coding rates under a joint excess-distortion probability constraint. The main contribution is an achievability proof based on a covariance shell covering lemma, analogous to the type covering lemma for discrete sources, and the asymptotic analysis is completed using the Berry-Esseen theorem.

Understanding the Value of Anarchy: Empirical Studies of Best Response Dynamics

2026-10-28T00:00:00+09:00

Toward Understanding the Value of Anarchy: Empirical Studies on Best Response Dynamics arXiv · Oct 28, 2026This paper experimentally tests best response dynamics (BRD) on randomly generated weighted networked multiagent optimization problems, comparing results with the Price of Anarchy (PoA) and with randomly selecting an equilibrium. The results show that simple best response mechanics tend to outperform both the PoA prediction and uniform random equilibrium selection, even when agents cannot observe all other agents' actions. The study also finds that when multiple equilibria exist, BRD appears biased toward higher-performing ones.

A comparative study of parallel short-time Fourier transform implementations

2026-10-28T00:00:00+09:00

A Comparative Study of Parallel Implementations of Short-Time Fourier Transform arXiv · Oct 28, 2026This paper compares two parallel algorithms for the short-time Fourier transform (STFT), which shows how a signal's frequencies change over time. One is FFT-based and the other is a quasi-parallel feedforward algorithm. Each is tested with shared-memory, distributed-memory, and hybrid parallel approaches, and the study analyzes runtime, memory use, and communication cost.

Improved Johnson Bound for List-Decoding with Hints

2026-10-28T00:00:00+09:00

An Improved Johnson Bound for List-Decoding with Hints arXiv · Oct 28, 2026This paper studies whether list-decoding radius for error-correcting codes can exceed the classical Johnson radius when some nearby codewords, called hints, are known. It determines the optimal list-decoding radius for single and multiple hints and shows it equals the optimum of a natural convex program. The authors also give an efficient weighted Guruswami-Sudan based algorithm that list-decodes Reed-Solomon codes up to this optimal radius.

Lower bounds on b-symbol weights of dual Melas codes and a Poisson randomness postulate

2026-10-28T00:00:00+09:00

On the b-Symbol Weight Distribution of Dual Melas Codes and a Poisson Randomness Postulate arXiv · Oct 28, 2026This paper proves a new lower bound on the b-symbol weight of codewords in dual Melas codes that is strictly better than known results. It recasts a pattern counting problem in linear-feedback shift register sequences as a problem about orbits of binary polynomials under a certain action. The same techniques yield an improved burst-covering radius bound for Melas codes and a new upper bound on the critical exponent of dual Melas codes. The authors also define a randomness postulate for binary sequences based on infinite Poisson generic sequences, conjecture that dual Melas codewords satisfy it, and report partial progress.

A Poisson transition and bias law for KSG mutual-information estimation

2026-10-28T00:00:00+09:00

A Poisson Transition and Bias Law for KSG Mutual-Information Estimation arXiv · Oct 28, 2026This paper studies the error of the KSG-1 mutual-information estimator on a Gaussian channel as sample size grows and the channel becomes nearly deterministic. The authors define a resolution parameter u_n = n·e^(-I_n) and prove that, when it converges to a positive limit, the estimation error converges in mean square to a computable deterministic bias. The bias is derived from an explicit marked-Poisson law for neighborhood counts, which also gives the limiting saturation probability.

Joint Trajectory and Resource Optimization for Secure UAV Networks Using Deep Reinforcement Learning

2026-10-28T00:00:00+09:00

Efficient Joint Trajectory Optimization and Resource Allocation for Secure UAV Networks: A Deep Reinforcement Learning Approach arXiv · Oct 28, 2026This paper studies how to jointly optimize UAV trajectories and resource allocation in wireless networks where a tethered UAV flying center and a malicious ground eavesdropper are present. The UAVs harvest energy and adaptively switch between transmitting, jamming, and relaying roles to deliver data to the tethered center. The authors formulate a utility maximization problem that balances secrecy rate, energy consumption, and interference, and solve it with a Proximal Policy Optimization based deep reinforcement learning algorithm. Simulations show better secrecy performance, lower energy use, and reduced interference than conventional and other DRL-based methods.

ConfluxServe: Balancing Resource Use and Parallel Efficiency in LLM Prefill and Decode

2026-10-28T00:00:00+09:00

ConfluxServe: Reconciling Resource Utilization and Parallelism Efficiency across Prefill and Decode in LLM Serving arXiv · Oct 28, 2026Prefill and decode use resources in complementary ways but prefer different parallelism strategies. The paper proposes Conflux Parallelism, which runs both phases in the same batch with different parallelism across operators. Building on this, it presents ConfluxServe, an SLO-aware serving system that maximizes goodput under TTFT and TBT constraints. On production workloads, it improves serving goodput by 1.26 to 1.62 times over the strongest baselines.

Interference-Conjugate Transform: A Tightest Surrogate for Wireless Power Control, Scheduling and Beamforming

2026-10-28T00:00:00+09:00

The Interference-Conjugate Transform: The Tightest Interference-Linear Surrogate for Power Control, Scheduling and Beamforming in Wireless Networks arXiv · Oct 28, 2026The paper proposes the interference-conjugate transform (ICT) for maximizing weighted sum rate in wireless networks. It proves that ICT is the unique best member of the family of interference-linear surrogates that includes FP, WMMSE, SEFP and XMMSE, and is pointwise dominant over these benchmarks. In uplink scheduling experiments, ICT achieves higher proportional-fairness utility than FP and SEFP using fewer iterations per slot, and in multi-cell MISO beamforming it converges in fewer iterations than the benchmarks.

Pooled testing under congestion: scaling laws for choosing pool sizes

2026-10-27T00:00:00+09:00

Pooled Testing under Congestion: Operational Scaling Laws for Pool-Size Selection arXiv · Oct 27, 2026The paper models pooled testing with random sample arrivals and limited testing capacity as a queueing system. It derives exact performance formulas and asymptotic scaling laws for fixed-batch Dorfman testing and a dynamic two-threshold policy. The optimal pool size depends jointly on prevalence, capacity, and delay, following a square-root law, a cube-root law, or remaining bounded depending on the regime.

Minimum subpacketization of Reed-Solomon codes for cut-set-optimal linear repair

2026-10-28T00:00:00+09:00

The Minimum Subpacketization of Reed-Solomon Codes with Linear Repair Achieving the Cut-Set Bound arXiv · Oct 28, 2026This paper studies the smallest subpacketization of Reed-Solomon codes that allow exact linear repair of every coordinate from any d helpers while meeting the cut-set bound. It proves a lower bound of sG_s(d+1) and shows it equals sG_s(n) when d(d+2) ≥ 4(n+1). The paper also gives a finite formula for G_s(n), an asymptotic growth result, and an affine-plane example showing sG_s(n) is not universal, along with a conjecture. ★★★★ Notable: The authors show a lower bound sG_s(d+1) on the minimum subpacketization of RS codes with exact linear repair, and state that for d(d+2)≥4(n+1) it equals sG_s(n), covering full-helper repair for distr

Adapting hybrid LLM inference to each request for greater efficiency

2026-10-28T00:00:00+09:00

Request-Adaptive Execution for Efficient Hybrid LLM Inference arXiv · Oct 28, 2026This paper presents Morph, a serving system that co-designs model architecture and runtime so hybrid language models can choose their layer layout per request. Each block pairs Attention and state space model layers, and a prefill-stage gate selects a layout that stays fixed during decoding. Experiments on six datasets report a better quality-efficiency trade-off than static hybrid baselines, with serving throughput improved by up to 3.87x and substantially lower inference latency.

Privacy-preserving over-the-air federated learning in cell-free massive MIMO with user-centric clustering

2026-10-28T00:00:00+09:00

Privacy-Preserving Over-the-Air Federated Learning over Cell-Free Massive MIMO System with User-Centric Clustering arXiv · Oct 28, 2026This paper studies a cross-layer design for differentially private federated learning over cell-free massive MIMO networks, where user model updates are aggregated over the air. It shows that thermal noise in wireless reception can help mask private information, reducing or removing the need for artificial noise. The authors jointly optimize access point-user associations, transmit powers, and artificial noise variances to minimize aggregation variance under per-user privacy constraints. Simulations report up to 89% lower aggregation variance than baselines and 97% test accuracy on MNIST, with consistent gains on CIFAR-10.

Capacity of private information retrieval from coded storage with colluding, robust and Byzantine servers

2026-10-28T00:00:00+09:00

The Capacity of Private Information Retrieval from Coded Storage Systems with Colluding, Robust and Byzantine Servers arXiv · Oct 28, 2026This paper studies private information retrieval from MDS-coded databases that include T colluding servers, U robust servers, and B Byzantine servers. The authors characterize the capacity for linear, full support-rank schemes and show the same expression holds without those restrictions when K=1 or T=1. They also extend the construction to retrieving P out of M files, and report that for P≥M/2 with K=1 or T=1 the multi-file scheme achieves capacity.

Uniqueness of codes for 4-cell connected disjoint clusters in the spatial chromatic model

2026-10-28T00:00:00+09:00

Code uniqueness of 4-cell connected disjoint clusters in spatial chromatic model arXiv · Oct 28, 2026This paper studies the coding of the spatial chromatic model (SCM), an entity-oriented spatial data model in which each cell has a chromatic code. Earlier work had shown uniqueness for 2-cell and 3-cell connected disjoint clusters. Building on that, the authors prove that the chromatic codes of 4-cell connected disjoint clusters are unique.

TAO: Tool-Aware Joint Scheduling for LLM Agent Serving

2026-10-27T00:00:00+09:00

TAO: Tool-Aware Joint Scheduling for LLM-Based Agentic Serving arXiv · Oct 27, 2026This paper presents TAO, a scheduling framework for serving LLM-based agents that alternate between GPU model inference and external tools running on CPUs. TAO's value function combines near-term tool readiness, node-specific cache credit with time-aware decay, and token footprint, and a dynamic program jointly selects and places waiting sessions under capacity limits. Across 12 reported model-workload-concurrency settings, TAO reduces latency by up to 26.8%, mean time to first token by up to 36.3%, and increases decode throughput by up to 29.5% relative to the best evaluated baseline for each metric.

PipeLink: Contract-Guided Linking of GPU Kernel Components

2026-10-27T00:00:00+09:00

PipeLink: Contract-Guided Linking of GPU Kernel Components arXiv · Oct 27, 2026This paper presents PipeLink, a linker that combines separately compiled GPU kernel components using contracts stating what each component requires and provides. PipeLink checks these contracts, solves a joint placement of streaming multiprocessor resources, and generates the transfers between components. Across 51 configurations on Hopper and Blackwell GPUs, linked kernels run within 0.2% of the same sources compiled as one unit. ★★★★ Notable: PipeLink links separately compiled GPU kernel components via contracts stating their requirements and promises, solving one joint placement of SM resources and generating inter-component transfers. Th

Negative Results for Over-the-Air Split Learning Under Channel Aging

2026-10-27T00:00:00+09:00

Negative Results for Reciprocity-Based Over-the-Air Split Learning Under Channel Aging: Stability, Alignment, and Geometry arXiv · Oct 27, 2026This paper studies reciprocity-based over-the-air split learning, where channel aging rotates and attenuates the derivative signal sent back to the transmitter. Seven case studies combine numerical forecasts, mechanism checks, and competing predictions. Most predicted training accuracy gains are not supported, while some local physical relations, such as the inverse relation between amplitude and phase gradient, hold.

Deep Holes and Aperture: Sign Laws for Lattice-Sampled Random Waves

2026-10-27T00:00:00+09:00

Deep Holes and Aperture: Sign Laws for the Empirical Correlation of Lattice-Sampled Random Wave Fields arXiv · Oct 27, 2026This paper studies whether the empirical lag correlation of a Gaussian random wave field, sampled on a lattice within a bounded aperture, keeps a positive real part throughout the Voronoi cell. It identifies deep holes as the minimizers that govern failure and gives explicit sign-class counts for square, hexagonal, simple cubic, and body-centered cubic lattices. The authors also derive asymptotic and non-asymptotic aperture laws and state when a failure floor appears under Rayleigh-faded specular paths.

Counting the fixed points of binary belief propagation: generic degree, syzygies, and real solutions

2026-10-27T00:00:00+09:00

Counting the Fixed Points of Binary Belief Propagation: Generic Degree, Syzygies, and Real Solutions arXiv · Oct 27, 2026The paper studies how many fixed points belief propagation has on a binary pairwise model. Over complex parameters, it determines the generic degree of the algebraic family arising from finite loopless multigraphs, expressed as a graph state sum over edge sets without degree-one vertices. Two geometric realizations of this state sum are given, and at strong external fields the real fixed point count is analyzed as a lower bound. The degree, resolution, and real-count results are formalized in Lean 4.

Wireless localization using digital twins and probabilistic inverse learning

2026-10-27T00:00:00+09:00

Digital Twin-Guided Wireless Localization via Probabilistic Inverse Learning arXiv · Oct 27, 2026This paper studies how site-specific digital twins can support wireless localization with simplified distributed infrastructure. It proposes PRISM, a probabilistic scheme using single-antenna access points, which learns a proposal distribution over user locations from channel impulse responses and reweights candidates using digital twin consistency. With 10 single-antenna access points, it reports a median error of about 0.7 m and remains effective without retraining when digital twin features are degraded.

Contention-aware scheduling to handle skewed sparse matrix multiplication (technical report)

2026-10-27T00:00:00+09:00

Taming Multi-Dimensional Skew in Sparse Matrix Multiplication with Contention-Aware Scheduling [Technical Report] arXiv · Oct 27, 2026This report examines why sparse matrix-matrix multiplication (SpGEMM) slows down on skewed real-world workloads. It identifies two causes: imbalance across computation cost, output density, and data locality, and cache contention from dense substructures. The authors propose adaptive morsel scheduling (AMS), which combines contention-aware load balancing, a data-conscious SMT activator, and a per-morsel hybrid accumulator. Evaluations on matrix squaring, relational queries, and graph analytics report that AMS significantly outperforms existing methods on skewed workloads while staying comparable on non-skewed ones.

Physics-informed neural operator for channel estimation in fluid antenna systems

2026-10-27T00:00:00+09:00

Rethinking Channel Estimation for Fluid Antenna Systems: A Physics-Informed Operator Learning arXiv · Oct 27, 2026This paper frames channel estimation for fluid antenna systems as a stochastic inverse field-reconstruction problem constrained by the projected Helmholtz model. It shows that finite-aperture reconstruction leads to a frequency-domain deconvolution problem and proposes a U-shaped Fourier neural operator trained with a physics-aware Sobolev loss. Simulations report higher accuracy than state-of-the-art benchmarks and much lower inference latency than iterative solvers at high spatial-frequency resolutions.

How the power delay profile threshold affects RMS delay spread estimates at 33 GHz

2026-10-27T00:00:00+09:00

Sensitivity of RMS Delay Spread Estimation to Power Delay Profile Thresholding in Time-Varying Channels at 33 GHz arXiv · Oct 27, 2026This paper examines how estimates of RMS delay spread, a measure of signal time dispersion in radio channels, change with the relative power threshold applied to power delay profiles. The analysis uses wideband measurements at 33 GHz in a dynamic road environment with moving vehicles. Lowering the threshold from -20 dB to -50 dB raises the median from about 140 ns to 168 ns and the 90th percentile from 200 ns to 210 ns. Both statistics converge at lower thresholds, with -40 dB a practical convergence level.

When Do Token Systems Stabilize? Compatibility Constraints and Community Structure

2026-10-27T00:00:00+09:00

When do token systems stabilize? Compatibility constraints and community structure arXiv · Oct 27, 2026This paper studies token systems that enable service exchange without money, where agents face compatibility constraints. It establishes a necessary and sufficient condition for stability and shows that a minimum-token rule, selecting the lowest-balance available provider, achieves the optimal stability region. The criterion is applied to random token systems generated by stochastic block models, identifying two obstructions to stability: insufficient service opportunities for a group of communities or for an individual agent.

Finite Fields: root sets, Frobenius maps and algebraic structure (monograph)

2026-10-27T00:00:00+09:00

Finite Fields: Root Sets, Frobenius, and Algebraic Structure arXiv · Oct 27, 2026This monograph develops finite field theory through root sets, Frobenius orbits, trace duality and cyclic modules. It applies these tools to polynomial constructions and algebraic coding theory, covering cyclic and constacyclic codes, BCH and Goppa bounds, rank-metric and trace codes. It also treats character sums, Gauss and Jacobi sums, and proofs of Weil bounds, emphasizing explicit links between field, orbit and linear-dependence structure.

Characterizing Maximum Nash Welfare: Envy-Freeness and Proportionality

2026-10-27T00:00:00+09:00

Characterizing Maximum Nash Welfare: Envy-Freeness and Proportionality arXiv · Oct 27, 2026This paper studies how to allocate indivisible goods using the maximum Nash welfare rule. This rule was already known to be the only welfarist rule satisfying envy-freeness up to one good. The authors show that this fairness requirement can be replaced by much weaker conditions for any fixed k: envy-freeness up to k goods, or proportionality up to k goods. The characterization also holds for preference profiles with identical goods.

Exact weight spectrum of RM(m-7,m): a counterexample to the Carlet-Solé conjecture

2026-10-27T00:00:00+09:00

The Exact Weight Spectrum of RM(m-7,m): A Counterexample to the Carlet-Solé Conjecture arXiv · Oct 27, 2026This paper studies the weight spectrum of the Reed-Muller code RM(m-7,m), which describes which weights occur among Boolean functions of degree at most m-7. Combining affine geometry and Fourier analysis, the authors exclude eight weights previously conjectured for RM(7,14). They extend these exclusions to all m≥14 and thereby provide a counterexample to the Carlet-Solé conjecture.

A new MIMO design using movable antennas that keeps signal paths strong

2026-10-27T00:00:00+09:00

Design of DoF-Robust MIMO via Generalized Pinching Antennas arXiv · Oct 27, 2026This paper examines how distance differences between users and a base station can cause multi-user MIMO channels to lose spatial degrees of freedom. It proposes generalized pinching antennas placed close to users, which make the channel Gram matrix nearly diagonal and well-conditioned. Performance is assessed using rank-deficiency and outage probabilities. The analytical and simulation results indicate that pinching antennas can preserve spatial degrees of freedom and support users at spatially separated locations.

Agar: What-If Analysis for Microservice Applications

2026-10-27T00:00:00+09:00

Agar: What-If Analysis for Microservice Applications arXiv · Oct 27, 2026This paper presents AGAR, a counterfactual performance analysis tool for microservice applications. Instead of predicting exact values for each component, it learns simple, interpretable bounds that contain all observed behaviors and are tight enough to support end-to-end conclusions, then composes them to answer SLO-feasibility questions. This framing turns autoscaling into the query of finding the smallest resource allocation that meets an SLO. Across three microservice applications, the derived autoscaler reduces CPU allocation by up to 36.8% compared with state-of-the-art horizontal autoscalers while satisfying application SLOs.

Noise Shapes the Surplus: Strategic Play Against No-Regret Learners

2026-10-27T00:00:00+09:00

Noise Shapes the Surplus: Strategic Play Against No-Regret Learners arXiv · Oct 27, 2026This paper studies how an optimizer in zero-sum games can gain cumulative payoff above the minimax benchmark against no-regret learners, here Follow-the-Regularized-Leader algorithms, that act on noisy payoff scores with decaying learning rates. Passive play earns at most constant surplus, while a clairvoyant optimizer secures surplus scaling with the inverse learning rate when the mean response is profitable. Active alternating play can still extract surplus from the learner's movement when it is not, and at the canonical rate the noiseless Ω(√T) surplus is recovered under mean-zero, finite-variance perturbations. ★★★★ Notable: The authors report that in zero-sum games, optimizers can still extract strategic surplus from noisy FTRL learners with decaying rates, including Ω(1/η_T) and Ω(∑η_t) results and an Ω(√T) rate at η_t∝

Optimal and Impossible Strategyproof Facility Placement in Multiple Dimensions

2026-10-27T00:00:00+09:00

Optimality and Impossibility in Multidimensional Strategyproof Facility Location arXiv · Oct 27, 2026This paper studies facility location in multidimensional normed spaces, aiming to minimize social cost, the sum of agents' distances to their nearest facility. For a single facility in the Euclidean plane, it shows that the optimal approximation ratios of deterministic strategyproof mechanisms match those of the coordinate-wise median, without assuming anonymity. It also shows that the coordinate-wise median is not always optimal in non-Euclidean spaces, and proves impossibility results for two or more facilities.

Learning-Augmented Coordination Mechanisms: Using Demand Predictions to Improve Equilibrium Efficiency in Congestion Games

2026-10-27T00:00:00+09:00

Learning-Augmented Coordination Mechanisms arXiv · Oct 27, 2026This paper studies how selfish behavior reduces efficiency in decentralized resource allocation. It proposes learning-augmented coordination mechanisms that use a possibly inaccurate demand prediction to modify resource costs, with a confidence parameter β that yields near-optimal performance when predictions are accurate while keeping a good worst-case equilibrium guarantee. The paper also gives robustness bounds and impossibility results, including limits on error tolerance in non-atomic games. ★★★★ Notable: Introduces learning-augmented coordination mechanisms for congestion games, aiming for near-optimal performance under accurate predictions while retaining PoS guarantees, with impossibility results.

JustGoal: An AI Given Only a Goal Beats Expert-Written GPU Kernels

2026-10-27T00:00:00+09:00

JustGoal: With Only a Goal, the AI Baseline Is Above the Expert Kernels arXiv · Oct 27, 2026The paper introduces JustGoal, a baseline in which a general-purpose AI agent receives only an optimization goal and constraints, without kernel-specific knowledge bases or predefined workflows. On NVIDIA B200 GPUs, it beats open-source libraries on 8 of 10 operators, with per-operator geometric mean speedups of 1.14x to 2.34x, and beats the fastest public specialized-agent kernel on 5 of 9 comparable operators. The authors conclude that specialized machinery should be justified by clear gains over a same-model, goal-only baseline and reevaluated when the foundation model changes.

Fundamental limits of bistatic ISAC over vector channels, i.e., the best achievable trade-off between communication rate and sensing accuracy.

2026-10-27T00:00:00+09:00

On the Fundamental Limits of Bistatic ISAC Over Vector Channels arXiv · Oct 27, 2026This paper studies the fundamental performance limits of bistatic integrated sensing and communication systems over vector channels. For parallel channels with a degradation structure, the authors establish matching achievability and converse results, giving a single-letter characterization of the capacity-distortion function. For vector Gaussian channels with additive Gaussian states, they prove that Gaussian superposition coding is optimal, yielding an exact characterization of the capacity-distortion-power function. ★★★★★ Notable: The authors study bistatic ISAC over vector channels, giving matching single-letter achievability and converse results for parallel channels. For vector Gaussian channels with Gaussian states, they re

Thermodynamic Characterization of Message-Passing Decoders: Framework and Initial Study

2026-10-27T00:00:00+09:00

Thermodynamic characterization of message-passing: Framework and initial study arXiv · Oct 27, 2026This paper studies the thermodynamic efficiency of message-passing decoders using the mismatch cost framework, which provides a computable lower bound on irrecoverable energy loss. Using the [7,4] Hamming code over a binary erasure channel, it shows how channel-induced initialization statistics and the step-wise evolution of the algorithmic state distribution determine the total mismatch cost. The results indicate that thermodynamic cost depends jointly on decoding dynamics, input statistics, and hardware characteristics, suggesting paths toward energy-efficient decoder design.

INC-MoE: Overlapping In-Network Collectives with Expert Computation for Faster MoE Inference

2026-10-27T00:00:00+09:00

INC-MoE: Overlapping In-Network Collectives and Expert Computation for Expert-Parallel Mixture-of-Experts Inference arXiv · Oct 27, 2026The paper presents INC-MoE, an expert-parallel mixture-of-experts layer that offloads token dispatch and output combining to the network switch and overlaps these collective operations with each other and with expert computation. In an emulated four-node deployment with a dedicated in-network computing switch, it was evaluated on OLMoE prefill with four-way expert parallelism. Relative to vLLM's default collective path, it reduces time to first token by up to 33% and raises throughput by up to 50% while matching outputs.

Converse of symplectic interlacing via low-rank singular value perturbations

2026-10-26T00:00:00+09:00

Converse symplectic interlacing via low-rank singular value perturbations arXiv · Oct 26, 2026The paper proves the converse of the symplectic interlacing theorem. Given a global symplectic spectrum and a local symplectic spectrum satisfying the 2k-step interlacing inequalities, it constructs a real symmetric positive definite matrix with these prescribed spectra, where the local spectrum belongs to a principal submatrix. The main tool reduces s-principal compression to a rank-k singular value perturbation. In the Gaussian quantum marginal problem, the symplectic interlacing inequalities are shown to be the complete spectral compatibility conditions for the global-versus-local-mode marginal problem. ★★★★ Notable: The authors prove the converse of the symplectic interlacing theorem, constructing a real symmetric positive definite matrix realizing given global and local symplectic spectra satisfying 2k-step inte

How learning algorithms pick equilibria and can lead to collusion

2026-10-27T00:00:00+09:00

How Learning Selects Equilibria and When It Leads to Collusion arXiv · Oct 27, 2026This paper studies how the design of learning algorithms affects which equilibrium players settle on in infinite-horizon discounted Markov games. It shows that stochastically stable states correspond to approximate stationary Nash equilibria that maximize the minimum transformed discounted value across players. In repeated Cournot games, every stochastically stable state yields an approximate public perfect equilibrium with minimum discounted profit above the Cournot-Nash benchmark, so collusion is inevitable under these conditions and learning dynamics. ★★★★★ Notable: The authors characterize equilibrium selection via stochastic stability and argue that learning dynamics lead to approximate collusion in repeated Cournot games, offering a new selection perspective o

Computational Hardness of Outplaying No-Regret Learners: Easy With Pure Play, Hard Even With Mixed Play

2026-10-27T00:00:00+09:00

The Complexity of Strategizing Against No-Regret Learners: Pure When It Is Easy, Hard Even When Mixed arXiv · Oct 27, 2026The paper studies how hard it is to optimize a strategy against a known online learner in repeated finite-horizon games. For general-sum games with bounded payoffs, unless P=NP, no polynomial-time optimizer can come within a constant factor cT of the best pure-action sequence, even when mixed strategies are allowed. In zero-sum games, deterministic pure actions efficiently recover surplus guarantees previously obtained only through mixed play against FTRL learners. ★★★★ Notable: The authors show that, under P≠NP, no polynomial-time optimizer against known learners in general-sum games can reach a constant fraction of the best pure-sequence reward even with mixed outputs, exte

Explicit Reed-Solomon codes with optimal list and line decoding

2026-10-27T00:00:00+09:00

Explicit Reed-Solomon Subspace Designs: Optimal List and Line Decoding arXiv · Oct 27, 2026The paper constructs explicit Reed-Solomon codes over extension fields that can be list-decoded from a fraction 1-R-ε of errors with optimal list size O((1-R)/ε). A simpler strongly explicit evaluation-point construction achieves list size (1/ε)^O(1/ε). Both constructions admit deterministic list-decoding algorithms running in polynomial time. ★★★★★ Notable: The authors give explicit Reed-Solomon codes with near-optimal constant list size via subspace designs, and report a new reduction yielding line decoding and proximity results.

Deterministic causal quantization using shared representatives with bounded error

2026-10-20T00:00:00+09:00

Deterministic Causal Quantization with Representative States and Bounded Error arXiv · Oct 20, 2026The paper studies deterministic causal quantization that keeps pathwise error at most ε, where shared representatives are updated using only transmitted addresses. It compares address entropy rates with the process ε-entropy and bounds the gap for stationary sources with finite scalar ε-entropy. The authors also present results for uniform lattice grids, autoregressive Laplace and Gaussian sources, and stationary processes on finitely many centers.

Re-examining fairness of pure Nash equilibria in Round-Robin allocation

2026-10-17T00:00:00+09:00

Revisiting the Fairness of Pure Nash Equilibria of Round-Robin arXiv · Oct 17, 2026The paper shows that, for additive valuations, not every pure Nash equilibrium of the Round-Robin mechanism yields an allocation that is envy-free up to one good (EF1) under agents' true valuations, correcting an earlier claim by Amanatidis et al. On the positive side, every pure Nash equilibrium of any fixed recursively balanced picking sequence mechanism is 1/2-EF1 and proportional up to one good (PROP1). The authors also give Round-Robin examples showing that no guarantee strictly better than 2/3-EF1 holds for all equilibria.

Optimal Test for Distribution Support Size

2026-10-14T00:00:00+09:00

Optimal Support Size Testing arXiv · Oct 14, 2026The paper presents an optimal tester that decides whether an unknown probability distribution is supported on n elements or is ε-far from such a distribution in total variation distance. Its sample complexity is Θ(n/(ε log n)), improving on the earlier O(n/(ε log n) · log(min{1/ε, n})) bound of Ferreira Pinto Jr. and Harms (STOC 2025). The authors also provide a fairly good lower bound on support size. ★★★★★ Notable: The authors present an optimal support-size tester with sample complexity Θ(n/(ε log n)), improving the prior logarithmic factor, and also give a lower bound on support size.

Logarithmic Separation Between LZ77 and LZ-End, and Grammar Compression

2026-10-24T00:00:00+09:00

Logarithmic Separation between LZ77 and LZ-End and Grammar Compression arXiv · Oct 24, 2026This paper proves a logarithmic separation between LZ77 and LZ-End. It also shows a logarithmic separation between LZ77 and grammar compression. These results establish that the compression efficiency of these methods differs by a logarithmic factor.

Scattering Measures and Information Geometry: Describing Scattering Experiments with Finite Measures and Their Information Structure via Fisher Metric

2026-09-28T00:00:00+09:00

Scattering Measures and Information Geometry arXiv · Sep 28, 2026The authors formulate scattering experiments as parameterized finite measures on outcome spaces and develop the information geometry defined by the Fisher metric. A scattering measure splits into total mass and normalized shape, separating Fisher information into changes in overall scattering strength and redistribution among outcomes. Measurement processes are modeled as probability or sub-probability kernels that contract Fisher information, linking the framework to optimal experimental design. The same finite-measure structure also arises in quantum measurements with incomplete detection.

Statistical Dark Matter: Synergy Is Everywhere but Hard to Detect

2026-09-28T00:00:00+09:00

Statistical Dark Matter: Synergy is everywhere but is hard to capture arXiv · Sep 28, 2026This paper studies synergy, an information-theoretic property found only in three or more variables jointly and absent from any subset of them. Using recent advances in multivariate information theory, the authors show synergy is far more common in complex systems than previously thought, and that it tends to dominate as the number of components grows. They also find that commonly used statistical modelling tools substantially underestimate these synergistic structures.

Calibrating a checksum detector for bfloat16 matrix multiplication errors and its blind spots

2026-09-29T00:00:00+09:00

Calibrating the Checksum: An Empirical False-Positive Bound for One-Sided ABFT on bfloat16 GEMM, and What It Fails to Catch arXiv · Sep 29, 2026This paper studies algorithm-based fault tolerance, a checksum method that flags corrupted matrix multiplications, which earlier work had evaluated only at float32. For bfloat16 with contraction length 4096, the authors fit a Generalized Pareto distribution to about 8e7 clean residuals and estimate a threshold near 3.14e-6 for a 1e-6 per-multiplication false-positive rate. Under their output-level perturbation model, the detector caught 304 of 400 single exponent-bit flips and none of 600 low-mantissa flips. The paper does not estimate real hardware fault rates.

A polynomial-time test for which graphs always admit strongly EFX orientations

2026-09-29T00:00:00+09:00

A Polynomial Time Characterization For Strongly EFX Orientable Graphs arXiv · Sep 29, 2026This paper studies fair division of discrete goods modeled as graphs, where vertices are agents and edges are goods. It gives a polynomial-time characterization of strongly EFX orientable graphs, which are graphs that admit an EFX orientation regardless of valuation. A connected graph qualifies if it is bipartite, or if its block decomposition has exactly one nonbipartite block containing a suitable degree-2 vertex whose removal leaves a bipartite graph. The authors state the proof was discovered by AI with human help in breaking the problem into subproblems. ★★★★ Notable: The authors give a polynomial-time characterization of strongly EFX orientable graphs, stating that a connected graph qualifies iff it is bipartite or has one nonbipartite block containing a degree-2

Simple deterministic auctions for selling multiple items

2026-09-29T00:00:00+09:00

On the Power of Determinism in Multi-Item Auctions arXiv · Sep 29, 2026This paper studies a single buyer with multiple items whose values are independent but not necessarily identically distributed. It analyzes the approximation ratios of three simple deterministic auctions: selling items separately, selling them as one bundle, and choosing the better of the two. For two i.i.d. items, the authors derive an exact tight ratio, and for general cases they prove that optimal revenue is at most 3 times the maximum of the two simple auctions' revenues, improving earlier bounds. ★★★★ Notable: The authors analyze three deterministic auctions for a single buyer with multiple items, report a tight 1+W(1/e)≈1.278 bound for two iid items, and prove REV ≤ 3·max{SREV,BREV}.

Geographic Incentives of Multiple Concurrent Proposers

2026-09-28T00:00:00+09:00

On the Geographic Incentives of Multiple Concurrent Proposers arXiv · Sep 28, 2026This paper studies whether systems with multiple concurrent proposers (MCP) incentivize proposers to decentralize geographically. Proposers strategically choose locations to capture user market share, since users pick where to send transactions based on time to inclusion. The authors formally prove that proposers' quorum radii, their distance to the nearest quorum of attesters, determine their incentives, so proposers minimize this radius. This yields co-location when proposers also act as attesters, but geographic diversity when those roles are separate. Pre-confirmations and faster intervalidator connections are shown to weaken the quorum radius effect and further destabilize co-location incentives.

Approximate uniformity in finite convolution models: how closely a sum of two fixed-support distributions can mimic the uniform law

2026-09-29T00:00:00+09:00

Approximate Uniformity in Finite Convolution Models arXiv · Sep 29, 2026This paper studies how well the sum of two independent distributions with fixed support can approximate the uniform law. It measures closeness using Manhattan distance, Euclidean distance, forward Kullback-Leibler divergence, and the line-metric Wasserstein-one distance. Explicit results are obtained for two models, including one conjecture. The authors also apply their reconstruction method to a recent conjecture about the coding capacity of an additive noise channel.

Surrogate-enhanced fractional programming for MIMO device-to-device interference management

2026-09-29T00:00:00+09:00

Surrogate-Enhanced Fractional Programming for MIMO Device-to-Device Interference Networks arXiv · Sep 29, 2026The paper addresses weighted sum-rate maximization in multi-stream MIMO device-to-device interference networks. It extends the scalar surrogate-enhanced fractional programming (SEFP) method to matrix ratios using Hermitian functional calculus, yielding a matrix SEFP framework. The authors show XMMSE is a special case of SEFP and propose SEFPLinQ for joint scheduling and beamforming, with numerical results reporting consistent gains.

Sunflowers of Reed–Solomon Codes: Families of Codes Sharing a Common Intersection

2026-09-27T00:00:00+09:00

Sunflowers of Reed--Solomon Codes arXiv · Sep 27, 2026The paper introduces Reed–Solomon sunflowers, families of Reed–Solomon codes whose pairwise intersections all equal the same fixed subspace. It gives an algebraic criterion using generalized V-matrices to identify such families. The authors study their sizes through counting and constructions, fully resolving the dimension-two case. For higher dimensions they give recursive and greedy constructions with many petals.

Heddle: Learning Structural Templates for Parallelism Planning on Heterogeneous GPU Clusters

2026-09-28T00:00:00+09:00

Heddle: Learning Structural Templates for Parallelism Planning on Heterogeneous GPU Clusters arXiv · Sep 28, 2026Heddle is a learning-based planner for training large machine learning models on heterogeneous GPU clusters. It reduces the planning problem to searching over pipeline structures, learning offline to build plans from structural templates across diverse cluster configurations. Evaluation reports that Heddle matches or exceeds the best of five existing planners, with throughput gains of up to 84.5% on dense models and 4.6x on MoE models.

Efficient Transmit Waveform Design for 1-Bit DAC MIMO-OFDM Radar-Communication Systems

2026-09-29T00:00:00+09:00

Efficient Transmit Waveform Design for MIMO-OFDM DFRC Systems with 1-Bit DACS arXiv · Sep 29, 2026This paper studies transmit waveform design for MIMO-OFDM dual-functional radar-communication systems that use 1-bit digital-to-analog converters, a topic it describes as underexplored. The waveform is designed to maximize the trace of the target angular Fisher information matrix, subject to symbol-error constraints via constructive interference and a 1-bit time-domain alphabet. The resulting large-scale nonlinear integer problem is solved with an inexact ADMM algorithm that alternates between time- and frequency-domain updates. Simulations show promising radar-communication performance.

HARS: An LDPC bit-flipping decoder that sets parameters from the initial syndrome and local reliability

2026-09-29T00:00:00+09:00

HARS: LDPC Bit-Flipping Decoding With Initial-Syndrome-Conditioned Parameter Mapping and Local-Reliability Weighting arXiv · Sep 29, 2026This paper presents HARS, a bit-flipping decoder for LDPC codes that combines local channel reliability with parameter selection based on the initial syndrome. The initial syndrome weight sets the base weight, reliability coefficient, threshold decay factor, and perturbation amplitude for each frame. On rate-one-half PEGReg and WiMAX codes, HARS reduces bit-error rate and mean iteration count compared with a calibrated SM-NGDBF baseline. An FPGA implementation on an XC7A200T is also reported.

A study of the maximum number of DNA sequences distinguishable by single-label DNA labeling

2026-09-29T00:00:00+09:00

On the Capacity of DNA Labeling in the Single-Label Setting arXiv · Sep 29, 2026DNA labeling distinguishes DNA sequences by recording where labels appear, and its labeling capacity is the growth rate of the number of distinguishable sequences as length goes to infinity. The authors model the labeling process as a deterministic channel and show its zero-error capacity equals the labeling capacity. For a single label this channel is a star graph, and they derive zero-error capacities for all star graphs, fully characterizing single-label cases. They also present a general method for constructing capacity-achieving codes.

Nonlinear data pricing makes optimal revenue computable again

2026-09-29T00:00:00+09:00

Non-Linear Pricing Restores Tractability for a Data Seller arXiv · Sep 29, 2026This paper studies how a data seller should price multiple datasets for budget-constrained buyers, using pricing functions that map purchased quantity to total payment. Prior work showed that optimal linear pricing is computationally intractable. The authors show that allowing general monotone, lower-continuous pricing functions can raise revenue and makes the optimum computable in polynomial time. Optimal functions are piecewise linear and convex, with total kinks bounded by the number of buyers.

Why Walker satellite routing tables are not periodic, and how to compare them with fiber

2026-09-29T00:00:00+09:00

On the Lack of Periodicity of Walker Satellite Constellation Routing Tables arXiv · Sep 29, 2026The paper models the configurations of a Walker satellite constellation as a dynamical system given by a translation on a torus, and extends this analysis to routing through satellites and fixed ground relays. For any deterministic, time-invariant routing rule, the routing table is periodic when the ratio of angular speeds is rational and non-periodic but ergodic when it is irrational. Long-run averages can then be computed from an explicit invariant measure, which makes comparisons of round-trip times between the constellation and terrestrial fiber meaningful.

ParaAnya: A plug-and-play output cache that speeds up parallel diffusion sampling by reusing repeated computations.

2026-09-29T00:00:00+09:00

ParaAnya: Accelerating Parallel Diffusion Sampling with Plug-and-Play Output Caching arXiv · Sep 29, 2026Diffusion model sampling is inherently sequential and computationally expensive. Parallel-in-time solvers use sliding windows, which causes the same timesteps to be evaluated repeatedly. ParaAnya caches input-output pairs and reuses them at overlapping timesteps, dispatching only cache misses to GPU workers. On Stable Diffusion v1.5 with eight GPUs, it gives 1.30 to 2.43 times speedups over uncached samplers and reduces function evaluations by up to 70.1% while keeping CLIP scores comparable.

Optimal rank-metric quantum codes and their exact recovery radii

2026-09-29T00:00:00+09:00

Singleton-Optimal Rank-Metric CSS Codes : Equality Structure and Exact Projected-Recovery Radii arXiv · Sep 29, 2026This paper studies rank-metric quantum codes for correlated faults that are dense physically but low-rank over a base field. It proves an asymmetric rank-metric Singleton bound for CSS codes, including degenerate ones, and characterizes equality via commuting maximum-rank-distance matrix codes. For Singleton-optimal pairs with positive logical dimension, the measured syndrome exactly determines the projected error modulo stabilizers at radii strictly below half the sector rank distance.

Selecting Contributors for Regional Digital Twins with Coverage Guarantees While Vehicles Move

2026-09-29T00:00:00+09:00

Conditional-Coverage Contributor Selection for Regional Digital Twins Under Mobility arXiv · Sep 29, 2026This paper studies how regional digital twins coordinate which contributors may join over congested broadcast networks in mobile settings without fixed infrastructure. The authors propose a host-coordinated protocol that admits only contributors whose timely, non-redundant coverage gain exceeds a congestion price within a cycle deadline, using a 13-byte beacon digest for pre-transmission scoring. Standards-based simulation reports that the timely actionable coverage deficit drops from 5.6% to 1.4% at comparable load to ETSI redundancy mitigation.

Algorithm for Arctic product-mix auction equilibria

2026-09-29T00:00:00+09:00

Solving the Arctic Product-Mix Auction arXiv · Sep 29, 2026This paper studies Arctic product-mix auctions, where budget-constrained bidders express preferences over multiple substitute goods. It gives the first strongly polynomial algorithm for finding competitive equilibria when the auctioneer has separable convex piecewise-linear costs. The method combines a strongly polynomial reduction to the costless Arctic auction with an existing strongly polynomial algorithm for that case. ★★★★ Notable: The authors give the first strongly polynomial algorithm for competitive equilibria of Arctic auctions with separable convex costs, plus reductions to linear Fisher and Arrow-Debreu markets.

Incentive-compatible online matching with random arrival order

2026-09-29T00:00:00+09:00

Arrival-Time Incentive Compatibility in Random Order Online Bipartite Matching arXiv · Sep 29, 2026The paper studies random-order online bipartite matching where users care only about being matched, while the platform maximizes total reward. It proposes the first constant-competitive, arrival-time incentive-compatible edge-weighted algorithm on complete bipartite graphs. The competitive ratio depends on the imbalance factor between online and offline nodes. For binary rewards, the authors present algorithms with a strictly better ratio.

LDPC Fractus Codes: Sparse Codes Built by Recursion

2026-09-29T00:00:00+09:00

LDPC Fractus Codes: Sparse Codes with Recursive Structure arXiv · Sep 29, 2026The paper introduces Fractus matrices, sparse matrices built through a self-similar recursive process, and uses them to construct LDPC codes. The resulting Tanner graphs have girth six and guarantee a minimum Hamming distance of at least ℓ+1. The authors establish algebraic properties of these matrices, including sparsity, regularity, recursive decomposition, and a natural lattice structure.

Influence maximization in near-linear time: a faster method for choosing key seed nodes

2026-09-29T00:00:00+09:00

Maximizing Social Influence in Almost Linear Time arXiv · Sep 29, 2026This paper studies influence maximization, which seeks k seed nodes in a graph with n nodes and m edges to maximize expected information spread under standard diffusion models. The prior Borgs et al. algorithm ran in Õ((n+m)k) time, and later work only reduced log factors or improved the dependence on ε. The authors present an algorithm achieving a 1−1/e−ε approximation in Õ(n+m) time, removing the multiplicative dependence on k from the running time. ★★★★★ Notable: The authors present an algorithm achieving a 1-1/e-ε approximation for influence maximization in near-linear time, removing the multiplicative dependence on k and answering a long-open question.

Voronoi-Markov chain and spatial entropy for point pattern analysis

2026-09-29T00:00:00+09:00

Voronoi-Markov chain and spatial entropy for point pattern analysis arXiv · Sep 29, 2026This paper revisits entropy in a spatial setting to derive computable, interpretable metrics for point pattern analysis, such as in histopathology. It reviews stochastic point process assumptions like Poisson homogeneity and Simple Sequential Inhibition, along with Voronoi and Delaunay tiling results that can provide null distributions for hypothesis testing. The authors introduce a "dense basin entropy" based on the Ord distribution over Voronoi basins, which is sensitive to clustering, and a Markov chain approximating Brownian motion. Simulations and histopathology experiments show these metrics offer insights complementary to other spatial entropy measures.

Fundamental Limits of Dynamic Tracking and Communication via Stationary Policies

2026-09-29T00:00:00+09:00

Fundamental Limits of Dynamic Tracking and Communication: A Stationary Policy Formulation arXiv · Sep 29, 2026This paper studies joint sensing and communication, where a multi-antenna transmitter communicates with a receiver while tracking a moving target. The sensing-communication trade-off is formulated as a rate-distortion problem, with a tractable outer bound on tracking error derived via the posterior Cramér–Rao bound. Under stated assumptions, stationary randomized Markov policies are shown to suffice, leading to a Gaussian covariance-control outer bound. Numerical experiments compare this outer frontier with an achievable frontier based on an extended Kalman filter.

DNN-Assisted Near-Optimal SCMA Codebooks for Rician Fading Uplink Channels

2026-09-29T00:00:00+09:00

DNN-Assisted Near-Optimal Codebook Design for Uplink SCMA Schemes over Rician Fading Channels arXiv · Sep 29, 2026This paper addresses Sparse Code Multiple Access (SCMA), a non-orthogonal multiplexing technique for next-generation networks, and proposes a deep neural network (DNN)-assisted method for designing near-optimal uplink codebooks. The DNN is trained unsupervised, using a loss function derived from codebook structural characteristics, so no labeled data is required. Rician-factor-dependent design criteria are derived, with AWGN and Rayleigh fading as limiting cases, along with upper and lower bounds on average mutual information. Numerical results show gains over representative benchmark schemes across a range of Rician factors and good error-rate performance in the error-floor region.

A hyperbolic bound on how fast two files can be retrieved from DNA storage

2026-09-29T00:00:00+09:00

A Hyperbolic Bound for File Retrieval in DNA-Based Data Storage arXiv · Sep 29, 2026The paper studies how sharing coding redundancy between two files affects their expected retrieval times in DNA-based storage. It focuses on two files encoded by a systematic linear code over a finite field. The authors prove a conjecture that the sum of each file's dimension divided by its expected retrieval time is at most one, provided one file has dimension greater than one. The proof uses a geometric view of growing spans and a sharpened projection bound.

Satellite semantic communication with vector quantization and index compression for high-resolution remote sensing images over AFDM

2026-09-28T00:00:00+09:00

Dual-Branch Vector-Quantization-Aided Satellite Digital Semantic Communication with Index Compression for High-Resolution RSI Over AFDM arXiv · Sep 28, 2026This paper proposes DVQ-SDSC, a dual-branch vector-quantization-aided satellite semantic communication framework for transmitting high-resolution remote sensing images over affine frequency division multiplexing (AFDM). It processes high-frequency residuals and low-frequency structural semantics separately, and reduces index bitrate using PCA-based codebook reordering and group DPCM (G-DPCM) coding. Experiments on FAIR1M over 3GPP NTN-TDL-D channels show it outperforms the JPEG-LDPC scheme at 0.0625 bits per pixel.

Sharing downstream profits with data sellers in tree-shaped data markets

2026-09-27T00:00:00+09:00

Profit Reallocation Mechanisms in Tree-based Data Trading arXiv · Sep 27, 2026Data can be copied and resold repeatedly, forming tree-shaped chains of resale, while upstream sellers capture none of the downstream value, which weakens their incentive to trade. The paper develops a framework for profit reallocation on such tree-structured markets, introducing a sequential trading game and a class of budget-feasible reallocation mechanisms, along with efficient algorithms for computing equilibria. The authors prove that any budget-feasible mechanism weakly expands equilibrium trades, and budget-balanced mechanisms also weakly improve social welfare. Experiments on synthetic markets show these benefits grow with the tree's depth and branching.

Competitive optimality of Bell-Cover randomization in testing by betting

2026-09-29T00:00:00+09:00

Competitive optimality in testing by betting via Bell-Cover randomization arXiv · Sep 29, 2026The paper extends the Bell-Cover randomization result to testing by betting for composite nulls and simple alternatives. It shows that the randomized numeraire e-variable UE* beats any other e-variable with probability at least one half. The authors prove this competitive optimality is equivalent to the numeraire inequality E_Q[E/E*] ≤ 1, and they discuss its costs in expected log wealth and power.

A truthful-in-expectation method for fairly splitting indivisible items, guaranteeing a constant share of each agent's maximin value

2026-09-29T00:00:00+09:00

Truthful-in-Expectation Mechanism with Constant Maximin-Share Guarantee arXiv · Sep 29, 2026The paper studies fairly allocating indivisible goods among n strategic agents with additive valuations. The authors give a randomized mechanism that is truthful in expectation and guarantees each agent at least 1/7 of her maximin share in every realized allocation. The mechanism is also ex-ante envy-free and runs in polynomial time. This confirms a conjecture that cardinal information allows a constant ex-post MMS guarantee. ★★★★ Notable: The authors present a truthful-in-expectation mechanism using cardinal information that guarantees each agent at least 1/7 of her MMS in every realized allocation, answering the conjecture of Babaioff

Dynamic wakeup when collisions are costly: a randomized algorithm

2026-09-29T00:00:00+09:00

Dynamic Wakeup under Costly Collisions arXiv · Sep 29, 2026This paper studies the wakeup problem on a shared communication channel, where packets become active at arbitrary times and a transmission succeeds only when exactly one packet sends. The authors propose Lowball, a randomized algorithm that needs neither collision detection nor knowledge of the number of packets. They measure performance by both latency and a per-collision cost C, and show that for large C the expected latency and expected collision cost are O(C^{1/2+ε} ln C) and O(√C).[ ]

A simple proof of the group-theoretic Zhang-Yeung inequality

2026-09-29T00:00:00+09:00

A simple proof of the group-theoretic Zhang-Yeung inequality arXiv · Sep 29, 2026This paper addresses the Zhang-Yeung inequality, the first non-Shannon-type inequality valid for all entropic vectors. Earlier work showed a one-to-one link between linear entropy inequalities and multiplicative inequalities on subgroup cardinalities of finite groups, but a direct group-theoretic proof of Zhang-Yeung was missing. The authors give a direct proof of this group-theoretic version using elementary group-theoretic and counting arguments.

Memory in the Sky: Drone Memories Combined to Answer Low-Altitude Questions

2026-09-29T00:00:00+09:00

Memory in the Sky: Low-Altitude Question Answering with Multi-Agent Memory Aggregation arXiv · Sep 29, 2026This paper studies low-altitude question answering (LAQA), where memories gathered by several drones are combined at a ground server to answer questions about long-term observations. It proposes a generative adversarial exam (GAE) to measure memory quality and a memory-centric framework (MemCen) that jointly selects drones and allocates transmit power under communication limits. MemCen reports QA accuracy of 92.4% and 84.0% in CARLA Town04 and Town05, and 88.5% on the PMAS real-world benchmark.

Binary k-center with a hard threshold, applied to delegated voting

2026-09-29T00:00:00+09:00

Binary k-Center under a Hard Threshold with Applications to Delegated Voting arXiv · Sep 29, 2026The paper studies covering binary strings of equal length, some possibly with missing entries, using a fixed number of center strings so that each input is within a given relative distance of its closest center. It is motivated by multi-issue approval voting, asking how representatives should be designed before an election so that as many voters as possible are willing to delegate. The authors give a complete picture of computational complexity across the number of centers and the agreement threshold when entries may be missing. For the case without missing entries, they extend the hardness results to any number of centers and large thresholds, leaving a narrow threshold range unresolved.

Online resource allocation where budgets can be refilled

2026-09-29T00:00:00+09:00

Online Resource Allocation with Replenishable Budgets arXiv · Sep 29, 2026Online resource allocation is a framework for sequential decisions under budget limits, usually assuming budgets can only decrease. This paper studies a more general setting where actions can either consume or replenish resources, motivated by inventory systems and energy markets. The authors propose a dual-based algorithm that recovers best-of-both-worlds guarantees in the standard case and improves them when the replenishment factor is positive. It achieves Õ(√T) regret in stochastic settings and Õ(√T) α-regret in adversarial settings, and strictly satisfies budget constraints.

Truthful-in-expectation fair-share allocation of chores with a constant guarantee

2026-09-28T00:00:00+09:00

Truthful-in-Expectation MMS Allocations for Chores arXiv · Sep 28, 2026This paper studies how to allocate indivisible chores fairly while keeping truthful reporting incentives in expectation. It presents the first such mechanism that guarantees a constant ex-post maximin share approximation for any number of agents, achieving 4/3 for two agents and 3/2 for three. For two agents, the authors raise the known lower bound to 4/3, showing their mechanism is optimal there. For any n of at least 3, they prove a lower bound of 13/12 on the achievable ratio.

Weaver: A System Using Idle GPU Capacity at AI-RAN Sites to Train Foundation Models

2026-09-28T00:00:00+09:00

Weaver: A System for AI-RAN Compute Sharing with Foundation Model Training arXiv · Sep 28, 2026The paper studies spare GPU capacity in AI-RAN cell sites, finding that 40-85% of GPU capacity goes unused, and that this spare capacity is complementary across sites. The authors present Weaver, which trains foundation models alongside radio access network workloads without degrading RAN performance, using a compute-aware MAC scheduler and elastic training. According to the paper, the prototype creates up to 4.9x more usable spare compute, and a multi-site testbed shows 2.1-3.7x higher training throughput than baselines.

Finite-Resolution MiLAC-Aided Multiuser Beamforming

2026-09-28T00:00:00+09:00

Finite-Resolution Microwave Linear Analog Computer (MiLAC)-Aided Multiuser Beamforming arXiv · Sep 28, 2026This paper studies multiuser beamforming with a microwave linear analog computer (MiLAC), whose tunable elements can take only finitely many states instead of continuous values. Under lossless and reciprocal constraints, it jointly optimizes discrete susceptances and power allocation, and designs codebooks from channel statistics. The authors propose exact continuous penalty models, proven globally equivalent to the discrete problems, and ADMM-based algorithms. Numerical results show a 3-bit stem-connected MiLAC reaches over 95% of the sum-rate of an unquantized fully-connected MiLAC using only 22% of its tunable components.

Envy-free splits of random assignments, with four agents solved

2026-09-28T00:00:00+09:00

Envy-Free Decompositions of Random Assignments: Settling Four Agents, and What Lies Beyond arXiv · Sep 28, 2026The paper studies whether a random assignment of objects to agents can be decomposed into deterministic assignments so that every agent envies any other with probability at most 1/2. It proves this is always possible for four agents using a computer-aided check over all vertices, and shows 1/2 is the best general threshold. It also gives proofs for special cases and shows some natural decomposition methods fail. Deciding whether an arbitrary random assignment admits such a decomposition is strongly NP-complete. ★★★★ Notable: The authors report that every four-agent SD-EF random assignment admits a Dec-EF decomposition via a computer-aided proof, and prove strong NP-completeness for the general decision problem.

AReaL-TIK: An agent that optimizes unified RL GPU kernels using a search-tracking representation

2026-09-28T00:00:00+09:00

AReaL-TIK: Stateful Agentic Optimization of Unified RL Kernels through an Optimization IR arXiv · Sep 28, 2026Reinforcement learning post-training often uses separate GPU kernels for rollout and policy updates, and numerical differences can distort probability ratios. The paper introduces KernelBraid, an agentic framework that starts from a hand-tuned, bitwise-consistent unified kernel and organizes code search with an optimization intermediate representation. Changes are kept only if they pass correctness checks and improve aggregate latency. Across 12 end-to-end training setups on H20, it reaches 1.10x average throughput relative to AReaL with log-probability recomputation. ★★★★ Notable: KernelBraid uses an optimization IR to guide agentic search for bitwise-consistent RL kernels; the authors report 1.10x average throughput over AReaL across 12 end-to-end configurations.

Does AI Advice Help or Harm Decisions? Study of Information Gathering with AI Input

2026-09-28T00:00:00+09:00

Does AI Help or Harm? Endogenous Information Acquisition with AI Advice arXiv · Sep 28, 2026This paper studies how AI advice given before a costly information-gathering step and final decision affects decision quality, measured by the gross utility of the final action. It characterizes gross effects through the value function generated by optimal acquisition under uniformly posterior-separable costs. The authors show that convexity of this value function is necessary and sufficient for every AI signal to weakly improve gross utility at every prior, while nonconvexity can allow harm. ★★★★ Notable: Proposes a convexity test on continuation value to determine when AI advice improves gross decision utility, with tight binary-state results and three- and four-state counterexamples showing possible

Tighter limits for constant-power and low-power error-correcting cooling codes

2026-09-28T00:00:00+09:00

Improved bounds for constant-power and low-power error-correcting cooling codes arXiv · Sep 28, 2026This paper studies binary constant-power and low-power error-correcting cooling codes for on-chip buses, which correct errors while controlling power and thermal effects, with e=w-3. The authors extend the earlier Zhao and Zhang upper bound for CPECC codes from w≥2t(t+1)+2 to w≥w0(t), where w0(t) is about √2·t^(3/2), and use Steiner systems to show this bound is attainable and asymptotically tight. For LPECC codes they give a new upper bound, ⌊C(n+2,3)/C(w+t,3)⌋, valid for w≥μ(t) with μ(t) about √2·t^(3/2), which is strictly smaller than the previous bound where both apply and is asymptotically tight.

Autonomous AI system finds new kissing number constructions in 19 dimensions

2026-09-28T00:00:00+09:00

Large-Scale Autonomous Discovery of Kissing Number Constructions arXiv · Sep 28, 2026The kissing-number problem asks how many non-overlapping spheres can touch a central sphere, and exact values are known only in a few dimensions. The authors used the autonomous multi-agent system Qiushi Engine to obtain new lower bounds in nineteen dimensions. Each result is supported by explicit mathematical arguments and independently checkable finite certificates. Examples include K(25)≥197580 and K(55)≥53301140. ★★★★ Notable: Using the autonomous multi-agent system Qiushi Engine, the authors report new kissing-number lower bounds in nineteen dimensions, from varied constructions with independently checkable finite certific

A proof of the generalized packing-covering conjecture (a coding-theory conjecture now proven)

2026-09-28T00:00:00+09:00

A proof of the generalized packing-covering conjecture arXiv · Sep 28, 2026This paper proves the generalized packing-covering conjecture for linear codes. The conjecture states that the t-th generalized Hamming weight is at most twice the t-th generalized covering radius plus two. The proof is computer-assisted, combining a parity-check reformulation, length bounds on possible counterexamples, and puncturing arguments. All orders t ≥ 32 are settled, and the remaining finitely many cases are excluded by exact computer verification. ★★★★ Notable: The authors give a computer-assisted proof of the generalized packing-covering conjecture, reporting that d_t(C) ≤ 2R_t(C)+2 holds for all linear codes and admissible orders t, combining parity-check

Monitoring Multitenant Kubernetes Clusters with TLA+ Trace Checking

2026-09-28T00:00:00+09:00

Monitoring and Verification of Multitenant Kubernetes Clusters using TLA+ Trace Checking arXiv · Sep 28, 2026This paper presents a monitoring tool that uses TLA+ trace-checking specifications at runtime, rather than only for design-time model checking. The tool infers system behavior from Kubernetes audit logs and detects violations in it. Its primary use case is multitenancy violations, though the pipeline is not limited to that setting. The work shows how formal reasoning can be incorporated into live Kubernetes environments to improve monitoring and correctness checking.

A method for synthesizing safe update schedules for safety-critical systems

2026-09-28T00:00:00+09:00

Synthesizing Update Schedules with Game-Based Extension of Bounded Model Checking arXiv · Sep 28, 2026This paper addresses how to apply software updates to safety-critical systems without interrupting operation or using cold spare hardware. The authors model the system and the update as a two-player timed game and use a bounded SMT encoding to synthesize a fixed global-time update schedule. The schedule is claimed to guarantee safe and complete deployment under all admissible system behaviors, demonstrated on an autonomous driving trajectory planner.

WaveAlign: Cache-Aware Query-Row Scheduling for Faster Sparse Attention in Long-Video Generation

2026-09-28T00:00:00+09:00

WaveAlign: Cache-Aware Query-Row Scheduling for Sparse Attention in Long-Video Generation arXiv · Sep 28, 2026Long-video generation with diffusion transformers produces very long token sequences, so attention becomes the main inference bottleneck. The paper proposes WaveAlign, a lightweight framework that reorders query rows in dynamic sparse attention, grouping rows by similar key/value access patterns and sorting them within GPU waves to improve L2 cache hit ratio and reduce HBM traffic. Because only query and mask rows are permuted, sparse-attention semantics and existing kernels are unchanged. Reported results across two GPUs, two video DiTs, and four sparse-attention methods include up to 1.25x kernel and 1.17x end-to-end speedup without quality loss.

Divergence-minimizing distribution matching for parallel and memory channels

2026-09-28T00:00:00+09:00

Divergence-Minimizing Distribution Matching for Parallel Channels and Channels with Memory arXiv · Sep 28, 2026The paper extends distribution matcher theory to distributions with memory and shows that the divergence-minimizing matcher selects the highest-probability target sequences, as in the memoryless case. A divergence scaling law is generalized to distributions driven by innovation processes. The theory is applied to parallel additive white Gaussian noise channels using a modified enumerative sphere-shaping method with a weighted energy constraint. A three-channel example shows that joint shaping across channels substantially reduces rate loss at short blocklengths compared with product matchers, and simulations with probabilistic amplitude shaping and a 5G-NR LDPC code confirm the gains.

Robust beamforming for OTFS massive MIMO under uncertain channel information, with GPU acceleration

2026-09-28T00:00:00+09:00

GPU-Accelerated Robust Beamforming for OTFS Massive MIMO with CSI Uncertainty arXiv · Sep 28, 2026This paper studies robust beamforming for OTFS-based massive MIMO systems when channel state information is uncertain. The authors minimize total transmit power under a chance constraint that guarantees a set outage probability, and model the uncertain channel with support vector clustering as an asymmetric uncertainty set. They convert the problem into a semidefinite program and design a GPU-accelerated parallel ADMM algorithm. Simulations reportedly show lower transmit power than conventional schemes and a speedup from the GPU-accelerated ADMM.

Automatic self-reduction gives stronger randomized distributed lower bounds

2026-09-28T00:00:00+09:00

Distributed Lower Bounds via Automatic Self-Reduction arXiv · Sep 28, 2026This paper studies randomized lower bounds for graph problems in the distributed setting. The authors show that self-reduction is a special case of classical round elimination and introduce a new error measure under which round elimination yields bounds beyond ω(log log n). They identify a class of problems where this improvement applies black-box. The paper proves Ω(√log n) randomized lower bounds for maximal matching on regular 2-colored graphs, 1/k-integral matching, and maximal H-packing. ★★★★★ Notable: Recasts self-reduction as a special case of classical round elimination via a new error measure, yielding ω(log log n) randomized lower bounds and Ω(√log n) bounds for several graph problems.

Robust Resource Management for Satellite-Air-Ground Networks Using Neural Network Channel Uncertainty Learning

2026-09-28T00:00:00+09:00

Robust Resource Management for SAGIN using DNN-Driven Channel Uncertainty Learning arXiv · Sep 28, 2026This paper studies joint robust beamforming and resource allocation in space-air-ground integrated networks when channel state information is uncertain. It formulates a chance-constrained problem that minimizes total transmit power while meeting user quality-of-service requirements under a set outage probability. A deep neural network learns the uncertain channel as an asymmetric uncertainty set, and an adaptive iterative algorithm jointly optimizes resource allocation and beamforming. Simulations show the method achieves better energy efficiency and robust reliability than traditional robust and non-robust approaches.

Covert communication over large fluid antenna arrays: how channel hardening affects detection

2026-09-28T00:00:00+09:00

Covert Communications over Enormous Fluid Antenna Systems (E-FAS): Impact of Channel Hardening on Detection arXiv · Sep 28, 2026This paper studies covert communication over enormous fluid antenna systems (E-FAS), where guided surface wave propagation creates a channel with a random covariance scale. With a passive warden present, the optimal likelihood ratio test reduces to an energy detector, and the authors derive its false alarm and missed detection probabilities. Numerical results show that finite-mode E-FAS channels improve covertness over the fully hardened Rayleigh limit for finite observation intervals. The effect of channel hardening is asymmetric across the legitimate and warden links.

Computational complexity of multi-agent binary-action contract design, mapping approximation limits within gross-substitutes rewards

2026-09-28T00:00:00+09:00

Settling the Complexity Landscape of Multi-Agent Contracts with Binary Actions arXiv · Sep 28, 2026This paper studies how hard it is to compute optimal contracts when several agents each choose between two actions, focusing on gross-substitutes reward functions and related classes. The authors show the problem is APX-complete even for OXS rewards, and they give an EPTAS for weighted matroid rank functions, with no general randomized FPTAS. For partition matroid rank functions they give an FPTAS. For ultra reward functions, they prove that no polynomial-time randomized value-query algorithm achieves a 2^{o(n)} approximation in expectation. ★★★★ Notable: For multi-agent binary-action contracts, the authors map the approximability boundaries within gross substitutes: APX-completeness for OXS rewards, an EPTAS for weighted matroid rank functions, and an

Torch-PIM: Profile-guided automatic decisions on offloading PyTorch code to memory-near compute

2026-09-28T00:00:00+09:00

Torch-PIM: Automated Profile-Guided PIM Offloading for PyTorch arXiv · Sep 28, 2026This paper presents Torch-PIM, a compiler framework that uses profile-guided optimization to decide whether each parallel loop in PyTorch-generated code runs on the host CPU or on processing-in-memory (PIM) hardware. Each loop is assessed by its amount of work and its memory boundedness, using host profiling or information from MLIR. The authors report speedups over CPU-only execution of up to 8.6x on tensor operators, 2.9x on MLP, 4.4x on Attention, 5.1x on GPT-J-6B, and 3.6x on LLaMA-7B across PIM configurations of 32 to 128 cores.

Contraction of Rényi divergences for discrete channels (how Rényi divergences shrink through noisy channels)

2026-09-28T00:00:00+09:00

Contraction of Rényi Divergences for Discrete Channels arXiv · Sep 28, 2026The paper studies strong data-processing inequality (SDPI) constants for Rényi divergences on finite spaces, examining how they depend on the Rényi order. It proves monotonicity and convexity properties, identifies support restrictions on the relevant distributions, and gives bounds relating the constants to χ² and Hellinger divergences. The results are applied to local differential privacy and to finite-time convergence bounds for Markov chains. ★★★★ Notable: The authors systematically characterize Rényi-SDPI constants on finite spaces, establishing their monotonicity and convexity in the order α, support restrictions, and tensorisation bounds, with applic

Nash Equilibria in Pacing-Strategy Auctions: Complexity and Inefficiency

2026-09-28T00:00:00+09:00

Nash Equilibria in Auctions with Pacing Strategies: Complexity and Inefficiency arXiv · Sep 28, 2026The paper studies Auctions with Pacing Strategies (APS) games, where utility-maximizing bidders across many simultaneous first-price auctions each choose one pacing multiplier that scales their bids. It shows some instances have no approximate pure Nash equilibrium and that deciding whether an equilibrium exists is NP-complete in general but polynomial when the number of bidders or items is fixed. When equilibria exist, both the Price of Anarchy and the Price of Stability are exactly e/(e-1). ★★★★ Notable: Introduces APS games and reports instances with no approximate pure Nash equilibrium, NP-completeness of equilibrium existence (polynomial if bidders or items are fixed), and Price of Anarchy and Stab

Semantics, Workflows, and Infrastructure: How Production AI Agents Are Served at Scale

2026-09-28T00:00:00+09:00

Semantics, Workflows, and Infrastructure: Understanding Agent Serving at Production Scale arXiv · Sep 28, 2026This paper analyzes a two-week trace of 11.7 million requests from a large-scale production platform for general-purpose AI agents, supported by inference infrastructure with over 10,000 GPUs. It examines the platform at three connected levels: task initiation semantics, workflow execution patterns, and infrastructure serving demands. The measurements reveal highly skewed request volumes across sessions, rare execution overlap among sibling requests, and context reuse across task boundaries. The authors then discuss deployment implications and open problems for future agent serving systems.

How partition shape and rank density affect the speed of distributed quantum circuit simulations

2026-09-28T00:00:00+09:00

The Shape of Speed: Impacts of Partition Geometry and Rank Density in Distributed Quantum Circuit Simulations arXiv · Sep 28, 2026The paper evaluates 764 validated configurations of 39-qubit distributed quantum circuit simulations on 1,024 nodes of the Fugaku supercomputer. It finds that torus partition geometry dominates runtime, with flat partitions running 1.73 to 2.31 times slower than near-cubic ones despite identical data transfer. Rank density is a secondary factor, reducing runtime by about 11% at 16 ranks per node only on compact geometries.

VarioPath: Workload-Aware All-to-All Communication for PCIe GPU Clusters

2026-09-28T00:00:00+09:00

VarioPath: Workload-Aware All-to-All Communication for PCIe GPU Clusters arXiv · Sep 28, 2026This paper presents VarioPath, a scheduling framework for AlltoAllv communication on PCIe GPU systems, which is important for distributed mixture-of-experts model inference. It pairs an offline topology-aware analyzer with an online demand-aware scheduler to handle link contention and skewed, changing traffic. On four platforms with up to 256 GPUs, the authors report average AlltoAllv speedups of 5.88x over FAST and 1.72x over DeepEP. End-to-end, Qwen3 inference latency falls by up to 27.2% and Wan2.1 generation latency by 6.1%.

HyDra: a scalable system that fixes load imbalance in dynamic context parallelism for long-context training

2026-09-28T00:00:00+09:00

HyDra: Demystifying and Taming Dynamic Context Parallelism at Production Scale arXiv · Sep 28, 2026The paper studies dynamic context parallelism (DCP) for long-context model training, where Megatron-Core DCP sizes parallelism to fit memory, causing uneven computation across ranks. In a production 256K-context job on over 11K GPUs, per-rank microbatch times differed by up to 5x. The authors present HyDra, a load-driven scheduler combined with a nested Ulysses-in-ring CP engine. On a 2,048-GPU production job, HyDra reduced the pipeline bubble from 36% to 14% and improved throughput by 1.10 to 1.43x over Mcore DCP.

RIS-assisted multiuser NOMA networks robust to hardware impairments and imperfect channel information

2026-09-28T00:00:00+09:00

RIS-Assisted Multiuser NOMA Networks With Imperfect CSI Under Transceiver Hardware Impairments arXiv · Sep 28, 2026This paper studies RIS-assisted single-beam and multi-beam NOMA networks affected by transceiver hardware impairments and imperfect channel state information. The authors propose a robust scheme that minimizes base station transmit power while meeting users' target rates, solved with Bernstein-type inequality and semi-definite relaxation. Numerical results show the proposed scheme outperforms a non-robust RIS-assisted NOMA scheme that ignores these impairments, and multi-beam NOMA achieves higher beam gain than single-beam NOMA.

Offline-designed variances for parallel perturbed successive cancellation decoding of polar codes

2026-09-28T00:00:00+09:00

Parallel Successive Cancellation Perturbation-Enhanced Decoding of Polar Codes via Offline Variance Design arXiv · Sep 28, 2026This paper proposes OVD-PSCP, a parallel successive cancellation perturbation-enhanced decoder for short and medium polar codes, in which the perturbation variances are chosen offline before decoding. Because all branches, including the original SC branch, can start simultaneously, the method avoids the extra latency of schemes that generate perturbations from earlier decoding outcomes. Simulations for rate-1/2 polar codes of lengths 64 to 512 show lower block error rates than conventional SCP with the same number of perturbation branches, with larger gains for shorter codes and more branches.

Repairing DAG workflow schedules after link disruptions

2026-09-28T00:00:00+09:00

Schedule Repair for DAG Workflows under Link Disruptions arXiv · Sep 28, 2026This paper studies how to repair schedules for DAG workflows in networked IoT systems when link disruptions invalidate parts of them. It compares four repair scopes: waiting out the disruption, rerouting, rescheduling only affected tasks, and rescheduling all pending tasks globally, each measured against an oracle and charged a latency cost. Across 100 workload instances and different disruption processes, no single scope wins, and the authors conclude the repair scope should adapt to the disruption process and repair cost.

Provably efficient, manipulation-resistant linear regression: tight bounds

2026-09-28T00:00:00+09:00

Tight Efficiency Guarantees for Strategyproof Linear Regression arXiv · Sep 28, 2026The paper studies the trade-off between prediction accuracy and incentive compatibility in linear regression, where agents privately report labels and prefer predictions close to their true values. The authors give a deterministic group-strategyproof mechanism achieving a (d+1)-approximation to the least-squares optimum and show this is optimal even among universally strategyproof randomized mechanisms, answering an open question from Chen et al. (EC 2018). Under strategyproofness in expectation, squared individual loss keeps the factor d+1, while absolute individual loss achieves the tight ratio 2-1/(⌈d/2⌉+1). ★★★★★ Notable: The authors design a deterministic group-strategyproof mechanism for linear regression with a (d+1)-approximation, claim it is optimal even among universally strategyproof randomized mechanisms, and r

Performance vs Portability in HPC: Why Benchmark Before Running

2026-09-28T00:00:00+09:00

Performance vs Portability in Heterogeneous HPC Environments: Why Pre-execution Benchmarking is Required arXiv · Sep 28, 2026This paper compares running quantum chemistry software in Podman containers against running it directly on HPC systems, using precompiled executables with CPU-optimized builds across multiple compute nodes and no extra network tuning. The results show that performance depends on how the software was compiled, startup overhead, and execution configuration. Container preparation can dominate short tasks, while long calculations need progress monitoring. The authors conclude that workload-specific benchmarking and user-level monitoring are important when deploying scientific applications in containers on HPC systems.

Soft-Output GRAND-Aided Macrosymbol: A Joint Detection and Decoding Method for NOMA

2026-09-28T00:00:00+09:00

Soft-Output GRAND-Aided Macrosymbol arXiv · Sep 28, 2026This paper extends GRAND-AM, a coding-based joint multiuser detection and error correction method for NOMA systems, into a soft-output version called SOGRAND-AM. It produces calibrated bitwise a posteriori error probabilities and works with soft-decision outer forward error correction decoders. According to the paper, SOGRAND-AM achieves about a 2 dB bit error rate improvement at the outer soft-input decoder's output compared with hard-input GRAND-AM.

Proof of the Braun-Etzion-Vardy bound for binary subspace codes

2026-09-28T00:00:00+09:00

A proof of the Braun-Etzion-Vardy bound for binary subspace codes arXiv · Sep 28, 2026Linear subspace codes are codes defined within projective spaces. Braun, Etzion and Vardy (2013) conjectured that a linear subspace code in projective space has at most 2^n codewords. This paper resolves that conjecture. It uses a novel character-theoretic argument and the proof is self-contained. ★★★★★ Notable: The authors resolve the 2013 Braun-Etzion-Vardy conjecture that linear subspace codes in projective space have at most 2^n codewords, via a new character-theoretic proof they describe as self-containe

A continuous projection converse and weighted construction for uniquely decodable code pairs

2026-09-27T00:00:00+09:00

A Continuous Projection Converse and a Weighted Construction for Uniquely Decodable Code Pairs arXiv · Sep 27, 2026This paper studies the maximum sum rate of uniquely decodable code pairs. A weighted refinement of a complement-gluing construction gives an explicit code of length 672 with sum rate above 1.318639029203. An iterated projection argument, combined with a conditional entropy estimate and a justified continuous limit, yields an upper bound of 1.480063425539.

Randomized allocations balancing ex-ante and ex-post fairness under subadditive valuations

2026-09-27T00:00:00+09:00

On best of both worlds allocations with subadditive valuations arXiv · Sep 27, 2026This paper studies allocating indivisible goods to agents with equal entitlements and subadditive valuations. The authors give a transformation that turns any algorithm producing ρ-MMS allocations into a randomized algorithm that is ρ-MMS ex-post and η-MES ex-ante. They prove η is at least min[ρ/(2+ρ), 1/4] and apply the transformation to existing algorithms for subadditive and XOS valuations.

Adaptive client clustering and coordination for managing federated learning workflows at the edge

2026-09-27T00:00:00+09:00

Adaptive Client Clustering and Coordination for Federated Learning Workflow Management in Edge Networks arXiv · Sep 27, 2026This paper proposes A-CoDa, a framework for managing dependent federated learning workflows in edge networks. It groups clients by label-distribution divergence and selects participants using FedMIX, which ranks clients by loss, latency, and uncertainty. A dependency-aware DAG scheduler orchestrates layer-wise task execution. Experiments on handwriting, wearable-sensing, product-image, and medical-imaging tasks show reduced end-to-end completion time while maintaining competitive accuracy.

Gaps and Augmentations in Bayesian Scheduling

2026-09-27T00:00:00+09:00

Gaps and Augmentations in Bayesian Scheduling arXiv · Sep 27, 2026This paper studies how prior information, Bayesian incentive compatibility, randomization, and machine-side resource augmentation affect the approximation barrier for makespan scheduling on unrelated machines. It proves an asymptotic 4/3 lower bound for randomized BIC mechanisms with two machines and shows randomized BIC mechanisms can beat deterministic ones, with an integrality gap of 8/7. It also shows that under a spreading condition on job assignments, O(m/ε) sampled layers let the truthful MinWork mechanism reach a 1+ε approximation.

Toward system-of-systems integration for composable cloud, HPC, and edge AI platforms

2026-09-27T00:00:00+09:00

Toward System-of-Systems Integration for Composable Cloud-HPC-Edge AI Platforms arXiv · Sep 27, 2026This paper argues that modern AI platforms combining cloud, HPC, cloud-native, and edge stacks can be viewed as systems of systems, where each constituent keeps its own control and failure semantics. It proposes composable integration based on interfaces, contracts, and operational evidence, without requiring a single orchestration stack. The authors present a peer constituent-system view, a boundary test, a shared responsibility model, seven integration surfaces, and a cross-system workflow example. They close with evidence classes and research questions for evaluating interoperability, governance, observability, and fault containment.

Geometry-aware multi-UAV full-duplex communication system and field experiment

2026-09-27T00:00:00+09:00

Geometry-Aware Multi-UAV Full-Duplex Communication: System Design and Experiment arXiv · Sep 27, 2026This paper presents a multi-UAV in-band full-duplex communication system that uses high-gain directional antennas and separated uplink and downlink channels. It turns self-interference into controllable co-channel interference between UAVs, avoiding energy-intensive self-interference cancelers on the UAVs. The authors derive a geometry-aware interference model and define a reliable operating region in 3D airspace where the SINR requirement is met. Field trials with two UAVs and a ground station show downlink capacity that exceeds conventional TDD and approaches ideal full-duplex in most of the airspace.

Coded Computing for Dynamic Systems Using Cartesian Products

2026-09-30T00:00:00+09:00

Coded Computing for Dynamic System via Cartesian Products arXiv · Sep 30, 2026Coded distributed computing reduces Shuffle communication bottlenecks through coded multicast, but existing schemes assume a fixed set of workers and do not adapt to arrivals and disconnections. This paper proposes a dynamic scheme that keeps surviving workers' caches and Reduce assignments while using storage from newly arriving clusters. File placement uses Cartesian products of cluster-wise placement delivery arrays, and a greedy algorithm reassigns abandoned Reduce functions. The authors derive the exact communication load for arbitrary feasible disconnections and show, under one-shot communication and uncoded placement, that the scheme is within a factor of two of the lower bound when new clusters arrive without disconnections, and within a factor of four otherwise.

Codes that fix two deletions using only three extra symbols over large alphabets

2026-09-29T00:00:00+09:00

Systematic Codes Correcting Two Deletions with Three Redundant Symbols over Large Alphabets arXiv · Sep 29, 2026The paper builds systematic codes over polynomial-size alphabets that correct two deletions using three redundant symbols, with polynomial-time encoding and decoding. A cardinality bound by Liu and Xing shows three redundant symbols are necessary, so the construction is optimal in this respect. The authors also show existence of such codes over alphabets of size O(n^4) and give a polynomial-time construction for a single burst of two adjacent deletions.

Four New Binary Codes Correcting Single Errors

2026-10-17T00:00:00+09:00

Four New Single-Error-Correcting Binary Codes arXiv · Oct 17, 2026The paper constructs four binary codes: a (16, 3072, 3) code, a (17, 6144, 3) code, an (18, 12288, 3) code, and a (19, 24576, 3) code. These yield four new lower bounds on A(n,3) for n = 16 to 19. Each code is a union of cosets of a linear code, and the constructions were found by a computer search using a tabu search algorithm.

Robustness in Mechanism Design: A Worst-Case Review of Auction and Exchange Mechanisms

2026-10-01T00:00:00+09:00

Robustness in Mechanism Design arXiv · Oct 1, 2026This article reviews robust design and analysis of auctions, using the computer science paradigm that measures robustness by the worst-case ratio, over environments, between a robust auction's performance and an optimal auction's. It covers robust analysis of the first-price auction for welfare, prior-robust auction design for revenue, and prior-robust exchange mechanism design for gains from trade. It also shows that the revelation principle can lose performance in robust design, quantified by the revelation gap.

From Pilots to Production: Lessons on Multi-Site Federated AI for Science

2026-09-30T00:00:00+09:00

From Pilots to Production: Lessons in Cross-Institutional Federated Training and Artificial Intelligence for Science arXiv · Sep 30, 2026Many valuable scientific datasets cannot be centralized due to proprietary, export-control, classification, or data-sovereignty limits, so the model must move to the data. The paper synthesizes lessons from U.S. Department of Energy national laboratories, industry deployments, and open-source efforts on moving federated AI from pilots to production. It organizes these lessons around a five-domain readiness frame covering data architecture, privacy and security, trust and verification, governance, and operations. It also identifies open problems such as asynchronous federation, leakage auditing, verification standards, and harmonized data contracts.

Recovery sets and service rates of codes built by Plotkin-type construction

2026-10-02T00:00:00+09:00

Recovery Set Structures and Service Rates of Codes Obtained by the Plotkin-type Construction arXiv · Oct 2, 2026This paper studies service rate regions for binary codes used in distributed storage, where data can be recovered from several disjoint server groups. It shows how recovery sets of a Plotkin-type code arise from those of its underlying code, and uses recovery hypergraphs to bound service rates. The results are applied to first-order binary Reed-Muller codes, showing that service rates for R(1,m) can be derived recursively from the generator matrix of R(1,2).

Practical codes for computing Boolean functions over channels for decision tasks

2026-10-01T00:00:00+09:00

Task-Oriented Boolean Function Computation: Practical Code Constructions arXiv · Oct 1, 2026The paper proposes a practical Reed-Solomon-based code construction for Boolean function computation, where a receiver determines a function value unknown to the transmitter. For noiseless binary channels, it derives finite-blocklength worst-case error bounds and characterizes rate-reliability tradeoffs across function families. The construction is extended to noisy channels by concatenation with a conventional channel code, and numerical results report SNR coding gains over conventional transmission for specific tasks.

The Cost of Truthfulness in Fair Cloud Resource Sharing

2026-09-30T00:00:00+09:00

The Price of Strategyproofness in Fair Multi-Resource Allocation for Cloud Computing arXiv · Sep 30, 2026This paper studies how to fairly and truthfully divide several divisible resources among users with Leontief preferences, a setting motivated by cloud computing. It introduces a parametric mechanism family (ASF) and a Corrected Resource Balancing (CRB) mechanism, measuring efficiency with the fair-ratio benchmark. For two resources, CRB achieves fair-ratio at most 1+1/n, which is asymptotically optimal; for three or more resources, deterministic mechanisms satisfying sharing incentive and strategy-proofness have fair-ratio exactly m. ★★★★ Notable: Introduces ASF and CRB mechanisms; the authors report improving the two-resource fair-ratio from about 1.268 to 1+Θ(1/n), with a matching 1+Ω(1/n) lower bound.

Certified Splitting: Checking Safe Cut Points for Maximal-Munch Tokenization

2026-10-05T00:00:00+09:00

Certified Splitting: Decidability and Anchor Supply for Maximal-Munch Tokenization arXiv · Oct 5, 2026This paper studies how to certify byte windows in maximal-munch tokenization, where a window is a certified anchor if, in any fully tokenizable input containing it, the token covering its last byte starts at the window's origin. It gives an online method for literal vocabularies and an offline verifier for arbitrary regular token sets. The authors prove that cutting inputs at certified anchors and scanning chunks independently reproduces sequential segmentation, and they supply executable checks and measurements from shipped programs. ★★★★ Notable: Presents a decidable anchor-certification framework with theorems on chunked splitting, edit locality and differential auditing, each with an executable check, and cites measured outputs of chunked sc

Large-scale analysis of IO500 benchmark results for storage performance

2026-10-05T00:00:00+09:00

Repository-Scale Performance Characterization of the IO500 Benchmark arXiv · Oct 5, 2026This paper analyzes 294 post-reset IO500 submissions from 116 sites between 2019 and 2025 to study storage performance. It examines how individual benchmark phases relate to composite rankings. Phases measuring similar I/O behavior are strongly correlated, while bandwidth and metadata performance show weaker, scale-dependent relationships. Temporal and log-level analysis reveals information not visible in aggregate scores.

Fitting Fluid Antenna Channels with Extreme Value Distributions for Accurate Outage Evaluation

2026-10-04T00:00:00+09:00

A FAS Channel Fitting Strategy Using Extreme Value Distributions for Accurate Outage Performance Evaluation arXiv · Oct 4, 2026This paper proposes a channel fitting strategy for single-antenna fluid antenna systems (FAS) that focuses on accurately evaluating outage probability (OP). It introduces a combined extreme value distribution (CEVD), using a generalized Pareto distribution for the left tail and a generalized extreme value distribution for global behavior. A modified mean-square-error criterion with logarithmic-domain errors and uniform sampling on the log CDF is developed to emphasize tail accuracy. Simulations show improved OP evaluation accuracy in the ultra-low-OP regime.

Near-optimal coded caching that keeps both files and user demands private

2026-10-04T00:00:00+09:00

Order-Optimal Coded Caching With File and Demand Privacy arXiv · Oct 4, 2026This paper studies coded caching where both file contents and user requests stay private: each user gets only its requested file and learns nothing about other files or other users' requests. The authors propose a scheme whose worst-case delivery rate is at most 9/2 times optimal, for all user and file counts and feasible cache sizes. They also derive lower bounds on delivery rate and exact results at minimal memory and unit delivery rate. ★★★★ Notable: The paper proposes a coded caching scheme protecting both file and demand privacy, which the authors report achieves a worst-case delivery rate within 9/2 of the optimum, and derives matching-up-to-co

SyclKittens: A Tile Programming Model for Intel GPU Programmers and Coding Agents

2026-10-03T00:00:00+09:00

SyclKittens: A Tile Programming Model for Programmers and Coding Agents on Intel GPUs arXiv · Oct 3, 2026The paper presents SyclKittens, a hardware-aware tile programming model for Intel GPUs that encodes known-good methods such as matrix-engine operand layouts, L1 prefetching, and fabric communication. Under the same agent and budget, an agent using SyclKittens reaches 82.1% of Intel's oneDNN library on a GEMM kernel, compared with 57.5% using raw SYCL. The resulting kernel suite reaches about 96% of oneDNN in geometric mean across GEMM shapes and runs Llama-3.1-8B inference 1.59x faster than torch.compile on one GPU. ★★★★ Notable: The authors present SyclKittens, an Intel GPU tile model encoding known methods; they report GEMM at 82.1% of oneDNN and Llama-3.1-8B inference 1.59x faster than torch.compile.

One-shot beam tracking for stacked metasurface-assisted satellite broadcasting

2026-10-05T00:00:00+09:00

One-Shot Beam Tracking for Stacked Intelligent Metasurfaces-Assisted LEO Broadcasting arXiv · Oct 5, 2026This paper proposes a physics-informed one-shot beam tracking method for low-Earth-orbit satellite links assisted by a multilayer stacked intelligent metasurface. It predicts signal-quality changes from orbital motion and maps them to the metasurface phase settings through an analytical Jacobian, requiring only a single regularized correction. Numerical results show it keeps about 92% to 94% of the rate of iterative reoptimization while running hundreds of times faster.

Optimistic reliable broadcast with three-step latency

2026-10-05T00:00:00+09:00

Optimistic Reliable Broadcast in Three Steps arXiv · Oct 5, 2026This paper studies latency tradeoffs in signature-free asynchronous reliable broadcast that allow optimistic two-step delivery. The authors present two constructions that improve optimistic, honest-broadcaster, and totality delay bounds within specific ranges of parties and faults. They also prove a matching lower bound showing that, at maximal optimism, totality delay is at least three steps. The results indicate that maximal optimism costs one extra totality step compared with Bracha's protocol without increasing good-case latency. ★★★★★ Notable: The authors propose signature-free asynchronous reliable broadcast constructions achieving optimistic two-step delivery, with a lower bound proof for 3f+1≤n<4f and a (2,3,3) latency result.

Modulation and Integration in a Local Processor Receiving Two Input Sources

2026-10-05T00:00:00+09:00

Modulation and Integration in a local processor having input from two distinct sources arXiv · Oct 5, 2026This paper uses partial information decomposition to analyse how a local processor that receives two integrated input streams transmits information. The authors revise earlier conditions for contextual modulation and add a condition for signal integration, computing exact information-theoretic results with the Ibroja method. They test these conditions across several activation functions and apply them to noisy simulation data from a model layer 5b pyramidal cell, reporting evidence of basal-apical modulation and integration.

A simple algorithm that beats the 1/3 approximation barrier for Nash equilibria in two-player games

2026-10-05T00:00:00+09:00

A Simple Algorithm Breaking the 1/3 Barrier for Approximate Nash Equilibria in Bimatrix Games arXiv · Oct 5, 2026Computing Nash equilibria in bimatrix games is PPAD-complete, and this paper studies the strongest approximation guarantee achievable in polynomial time. The authors propose a simple, symmetric, deterministic polynomial-time algorithm with no tunable hyperparameters, achieving an approximation of about 0.30954+δ and breaking the long-standing 1/3 barrier. The method extends the optimization framework of Tsaknakis and Spirakis with building blocks from the authors' work on strategy mixing and automated approximation analysis. ★★★★★ Notable: The authors propose a simple deterministic polynomial-time algorithm with no tunable hyperparameters, reporting an approximation of about 0.30954+δ for bimatrix games that breaks the long-standing 1/3

Row-weight limits on redundant checks for peeling decoding of random LDPC codes

2026-10-04T00:00:00+09:00

Limits of Weight-Bounded Redundant Checks for Peeling Decoding of Random Regular LDPC Codes arXiv · Oct 4, 2026The paper studies peeling decoding on the binary erasure channel when redundant parity checks can be added freely but each check has a bounded row weight. For random simple binary regular LDPC codes, it considers the parity-check matrix formed by all nonzero dual codewords up to a given weight. Within an explicit logarithmic row-weight range, this full family of redundant checks has the same asymptotic normalized-residual threshold as peeling with the original LDPC matrix. ★★★★ Notable: In a logarithmic row-weight range, the authors report that all low-weight redundant checks keep the peeling threshold asymptotically equal to the original LDPC matrix, and that heavier checks covering

Tighter Upper Bound on the Ingleton Score for Four Random Variables

2026-10-04T00:00:00+09:00

Improved Bounds on the Ingleton Score of Four Random Variables arXiv · Oct 4, 2026This paper studies four finite discrete random variables and shows that the negative Ingleton expression divided by joint entropy is at most 19/124 (about 0.15323). This improves the previously reported bound of 3/19 by 11/2356. The proof uses a copy program with four copy steps and six auxiliary variables, verified by exact rational linear-programming certificates. The authors also show that two published inequalities imply the bound 5/32.

Secure multi-access coded caching via a lifting transformation

2026-10-04T00:00:00+09:00

Secure Multi-Access Coded Caching: A Lifting Approach arXiv · Oct 4, 2026The paper builds secure cyclic multi-access coded caching schemes by re-encoding the caches of a secure single-access scheme while reusing its multicast message. Each user reads L consecutive shared caches, recovers its requested file, and learns nothing about other files. Deterministic Pascal-window and randomized complement-dual encoders are given, attaining exact minimum per-block storage over a suitable finite field. The resulting memory-rate family has approximation guarantees against converse bounds, and the L=K-1 tradeoff is proved exactly.

HiNa-MoE: Fast, Non-Intrusive MoE Inference on CPUs with Matrix Engines

2026-10-04T00:00:00+09:00

HiNa-MoE: High-Performance, Non-Intrusive MoE Inference on CPUs with Matrix Engines arXiv · Oct 4, 2026This paper presents HiNa-MoE, an operator library that speeds up Mixture-of-Experts model inference on Intel AMX-equipped CPUs. It keeps expert weights in standard layouts, uses a simple page-interleaved policy for NUMA-aware work partitioning, and converts decode-phase matrix-vector operations into small matrix-matrix operations. The authors report speedups of up to 3.37x for FFN kernels and up to 2.09x for end-to-end inference over baselines, and state that it works plug-and-play with existing frameworks.

CommuteProp: Decoupled Training for Communication-Bound Split LLM Fine-Tuning

2026-10-04T00:00:00+09:00

CommuteProp: Decoupled Training for Communication Bound Split LLM Fine-Tuning arXiv · Oct 4, 2026This paper addresses inefficiency in privacy-preserving split-learning fine-tuning of large language models, where clients sit idle waiting for server-side gradients. It proposes CommuteProp, an asynchronous algorithm that splits training into two concurrent phases, a cross-block forward-backward pass and an in-block weight update, so computation overlaps with communication. The authors provide asynchronous error and convergence analysis and derive an NS preconditioner to reduce staleness-induced noise. Experiments report substantial throughput gains with accuracy comparable to synchronous methods, and the method can be plugged into existing privacy-enhancement techniques.

Secure device-to-device coded caching with limited-radius connections using lifting

2026-10-04T00:00:00+09:00

Secure D2D Coded Caching under Radius-Limited Connectivity via Lifting arXiv · Oct 4, 2026This paper studies secure coded caching in a device-to-device network where users on a cycle can only communicate with nearby users within a fixed radius. Each user must generate delivery signals from its cache and recover its requested file without learning any other file. The authors construct such schemes using an outer secret-sharing code and a secure local caching scheme, and they prove information-theoretic secrecy along with converse bounds on peak sum load. Within certain memory ranges, the schemes achieve optimal load or a constant-factor approximation.

You Only Convert Twice: Ozaki Scheme II for Repeated Tensor Mode Products

2026-10-04T00:00:00+09:00

You Only Convert Twice: Ozaki Scheme II for Chained Tensor Mode Products arXiv · Oct 4, 2026This paper proposes OzII-RescaleBE, which keeps intermediate tensors in residue form during chained tensor mode products in Ozaki scheme II. Inputs are converted to residues only once at the start and reconstructed only at the end, with an implementation on INT8 Tensor Cores using CuTe DSL. Across three datasets and chain depths 3 to 7, normwise relative errors stay near 1e-16 to 1e-15, but rise to about 1e-13 at depth 8. Throughput and memory use are compared against two GEMMul8 baselines, with gains at smaller mode sizes and a slight deficit at size 256.

Optimal colorings in unweighted Max-k-Cut games can still be unstable to group deviations

2026-10-04T00:00:00+09:00

Social Optimality Does Not Imply Coalitional Stability in Unweighted Max-k-Cut Games arXiv · Oct 4, 2026This paper studies Max-k-Cut games on unweighted graphs, where optimal colorings maximize social welfare. The authors disprove a conjecture by showing that for every k≥3 there is a finite, simple, connected graph with an optimal coloring that is not a strong equilibrium. A coalition can profitably deviate to another optimal coloring, benefiting all its members while outside players lose and social welfare is unchanged. The paper also gives an exact identity linking coalition gains, cut changes and internal conflicts, and bounds monochromatic edges remaining after strong deviations from general Nash equilibria.

Scalable distortion-aware clustering for fronthaul-limited cell-free MIMO networks

2026-10-04T00:00:00+09:00

Scalable Distortion-Aware Clustering for Fronthaul-Limited Cell-Free MIMO Networks arXiv · Oct 4, 2026This paper studies how to group access points in uplink cell-free MIMO networks with fronthaul limits, using maximum ratio combining. The optimization depends only on large-scale fading coefficients, and asymptotic analysis gives deterministic equivalent expressions for average network sum rate that account for quantization distortion. The authors exploit the structure of the resulting quadratic-over-linear objective to develop a polynomial-time scheme that reaches the global optimum. Numerical results show gains over improved existing methods and competitive performance against an exhaustive-search global optimum based on small-scale fading.

Formalizing Microwave Analog Computing in Lean

2026-10-04T00:00:00+09:00

Toward a Lean Formalization of Analog Computing with Microwaves arXiv · Oct 4, 2026This paper takes a first step toward formalizing microwave analog computing in Lean, a proof assistant. Microwave signals can perform linear transformations directly as they propagate through a network. The authors formalize components such as hybrid couplers and phase shifters, their series and parallel connections, and the networks they form. They formally prove a necessary and sufficient condition for the transformations these networks can compute, and that the discrete Fourier transform of any power-of-two size is implementable. All proofs are checked by the Lean kernel, and the code is publicly available.

A unified framework linking sensing, computation, and communication in over-the-air federated learning

2026-10-04T00:00:00+09:00

Learning OTA: A Unified Framework for Edge Sensing, Computation, and Communication arXiv · Oct 4, 2026This paper addresses the fact that federated learning training data must first be collected by heterogeneous edge sensors, and that collection quality and cost affect learning performance. It models sensing noise, energy consumption, latency, and over-the-air aggregation distortion in an OTA federated learning system, and derives a non-convex convergence bound. Based on this, the authors formulate a per-round optimization balancing energy, latency, and learning quality, solved with an alternating procedure. Numerical experiments show the benefit of jointly optimizing sensing, computation, and communication decisions.

Fluid antenna secure channel-driven constellation: a physical-layer security method for hiding data in channel points

2026-10-04T00:00:00+09:00

Fluid Antenna Secure Channel-driven Constellation (FASCC): a Physical Layer Security approach arXiv · Oct 4, 2026This paper proposes the FASCC framework for downlink communication, where information is carried by constellation points formed through fluid antenna port selection and phase adaptation. Two complementary schemes are developed depending on whether the eavesdropper's channel state information is available. Numerical results show significant secrecy gains in bit error rate and mutual-information secrecy, even when the eavesdropper is much closer to the base station than the legitimate user.

Approximate feedback Nash equilibria in constrained games via model predictive control with a provable error bound

2026-10-04T00:00:00+09:00

Approximate Feedback Nash Equilibria in Constrained Differential Games via Model Predictive Control with Upper Bound Guarantees arXiv · Oct 4, 2026This paper proposes a model predictive control method that approximates feedback Nash equilibrium trajectories in differential games by repeatedly solving finite-horizon open-loop games. An auxiliary-game formulation selects prediction horizons and terminal costs, and hard constraints are incorporated through a constrained open-loop game. For the unconstrained case, the authors derive an analytic upper bound on the state-trajectory deviation from the true equilibrium, and they compare the method numerically with baseline approaches.

Joint source-channel coding of Gaussian sources over block erasure channels, with nonasymptotic bounds and channel-uniform normal approximations

2026-09-30T00:00:00+09:00

Joint Source-Channel Coding of Gaussian Sources over Block Erasure Channels: Nonasymptotic Bounds and Channel-Uniform Normal Approximations arXiv · Sep 30, 2026The paper studies lossy transmission of a Gaussian memoryless source over memoryless, possibly nonstationary block erasure channels at finite blocklength, under an excess mean-squared distortion criterion. It derives computable nonasymptotic achievability and converse bounds and establishes matching third-order, channel-uniform normal approximations. Numerical evaluations compare these joint source-channel coding bounds and their common approximation with an optimized symmetrized separate source-channel coding benchmark.

Plotkin-Type Bounds for Function-Correcting Codes on Monomial Maps

2026-10-03T00:00:00+09:00

Plotkin-type Bound on Function-Correcting Codes for Monomial Mappings arXiv · Oct 3, 2026This paper studies function-correcting codes (FCCs), which give the receiver stronger protection for a function of the message than for the message itself. For arbitrary functions over finite fields under the Hamming metric, the authors derive a generalised Plotkin-type bound on FCC redundancy that reduces to the classical Plotkin bound for ordinary error-correcting codes. Explicit bounds are also obtained for linear functions, Hamming-weight functions, and Hamming-weight threshold functions. The paper further derives an explicit redundancy bound for monomial mappings f(x)=x^n over the field F_{q^k}.

Resource pattern design for V2X-ISAC under communication quality constraints

2026-10-03T00:00:00+09:00

QoS-Constrained Resource Pattern Design for V2X-ISAC Systems arXiv · Oct 3, 2026This paper proposes a method for selecting sensing resource patterns in vehicle-to-everything integrated sensing and communication (ISAC) systems, considering both communication reliability and sensing performance. Sensing is based on the Fisher information matrix for joint range and velocity estimation, with a minimum packet reception ratio imposed as a QoS constraint. A greedy selection algorithm with one-swap refinement is developed to avoid exhaustive search. Simulations show improved sensing over conventional patterns while meeting the PRR requirement, with performance close to the exhaustive-search optimum.

Timely persuasion by two opposing senders: does competition help the receiver?

2026-10-03T00:00:00+09:00

Timely Persuasion with Opposing Senders: When Does Competition Help the Receiver? arXiv · Oct 3, 2026Two senders with opposing preferences each send status updates about a binary continuous-time Markov chain under individual budgets. The receiver follows the reports only if doing so yields at least the long-run utility of a prior-based default estimate. The authors model the interaction as a Stackelberg game, characterize its pure-strategy equilibria in closed form, and compare receiver utility with a single sender and with a receiver-optimal reporting benchmark.

Recovering a hidden linear system and its starting states from partial observations.

2026-10-03T00:00:00+09:00

Identification and Recovery of Linear Dynamics from Partial Trajectories arXiv · Oct 3, 2026This paper studies how to jointly recover an unknown system matrix A and a low-rank initial-state matrix X0 from observations of the state trajectories A^t X0 that cover only some coordinates. It gives necessary identifiability conditions, lower bounds on the number of observations, and a rank-based principle for generic local identifiability. When trajectories from a basis of the initial-state range are fully observed, it provides necessary and sufficient rank conditions for global recovery with explicit formulas. Numerical experiments on synthetic and real data assess performance under different sampling rates and horizons.

Computing Equilibria in Integer Games with Shared Constraints

2026-10-03T00:00:00+09:00

Computing Equilibria in Integer Programming Games with Shared Constraints arXiv · Oct 3, 2026The paper presents a cutting-plane algorithm for computing pure Nash equilibria in finite integer games where deviations depend on opponents' choices due to shared constraints. Conditional equilibrium inequalities are embedded as lazy constraints in branch-and-cut to optimize over exact or approximate equilibria. Evaluated on 3,300 instances, the method finds a socially optimal equilibrium in nearly all cases and certifies nonexistence in 16.

OACM: An Optimistic Asynchronous Communication Model for Large-Scale Spiking Neural Network Simulation

2026-10-03T00:00:00+09:00

OACM: Optimistic Asynchronous Communication Model for Large-Scale SNN Simulation arXiv · Oct 3, 2026This paper proposes OACM, a communication model for simulating large-scale spiking neural networks across distributed computers. It combines optimistic hybrid simulation with rollbacks, asynchronous one-sided communication via the UNR library, and adaptive message aggregation and routing on a 2D-HyperX topology. On a high-performance computing cluster simulating a multi-area marmoset brain model with up to 174 nodes, the authors report up to 1.4x speedup over the original CORTEX simulator and over 22.6x speedup compared to NEST.

Learning in large populations with anticipatory payoffs, combining Bregman divergence with Willems and Lyapunov methods

2026-10-03T00:00:00+09:00

Learning in Large Populations With Anticipatory Payoffs: Bregman Meets Willems and Lyapunov arXiv · Oct 3, 2026The paper studies evolutionary Nash equilibrium learning in large populations whose payoffs come from dynamic mechanisms. Using counterclockwise dissipativity (CCW), it shows convergence for a broad class of well-behaved learning rules that align with payoffs. It then builds rule-independent Lyapunov functions from CCW storage functions, filling the gap where earlier results gave convergence but no Lyapunov stability. For a nonlinear anticipatory mechanism, the storage function is a Bregman divergence, and the resulting Lyapunov function establishes global asymptotic stability. A congestion game with a nonlinear congestion-growth toll illustrates the results.

BitIR: A Cross-Architecture Fault Injection Tool for Testing How Heterogeneous GPU Programs Handle Faults

2026-10-03T00:00:00+09:00

BitIR: Cross-Architecture Fault Injection for Resilience Analysis of Heterogeneous GPU Applications arXiv · Oct 3, 2026This paper presents BitIR, a framework that injects single-bit faults at the LLVM IR level so the same perturbation can be compared across NVIDIA, Intel, and AMD GPUs. The authors ran large injection campaigns on the Polaris, Aurora, and Frontier supercomputers using heterogeneous benchmarks. Outcomes were sorted into masked results, silent data corruptions, and failures. The results show that identical faults behave differently across vendors, so resilience depends on the backend.

Deterministic Stackelberg Strategies for Robotaxi Vehicle Assignment

2026-10-03T00:00:00+09:00

Deterministic Stackelberg Strategies in Robotaxi Assignment Games arXiv · Oct 3, 2026This paper studies competition between two robotaxi platforms, where an incumbent must dispatch its fleet to limit riders poached by a reactive entrant. The authors model this as a zero-sum assignment game and give an exact integer linear program, plus a polynomial-time approximation algorithm. They prove that computing or approximating the optimal payoff is hard in general, and provide an exact dynamic programming method for riders and vehicles on a line. Experiments on real ride-hailing data show the LP-based method performs well in practice, and simple heuristics are competitive.

SIC-based multi-way communication using reconfigurable intelligent surfaces

2026-10-03T00:00:00+09:00

RIS Aided Multi-Way Communication Systems based on SIC Decoding Techniques arXiv · Oct 3, 2026This paper studies multi-way communication systems aided by a reconfigurable intelligent surface (RIS) using QPSK, where multiple users exchange information by exploiting the RIS's full-duplex capability. Three successive interference cancellation (SIC) ordering methods are compared by bit error rate through Monte-Carlo simulations, and an extension to RIS elements with phase-dependent amplitudes is considered. Results show that with few RIS elements, SIC ordered by increasing mean square error performs best, while with more elements, the optimization-based ordering performs best.

A tighter upper bound on the size of recovering pairs

2026-10-03T00:00:00+09:00

An Improved Upper Bound for Recovering Pairs arXiv · Oct 3, 2026The paper proves that for any recovering pair of nonempty families of subsets of an n-element set, the product of the family sizes is at most (2.23949)^n. Previously, combining a result of Mond, Souza and Versteegen with the Fang-Huang bound gave (2.2499)^n. The proof combines two entropy inequalities through a convex combination. ★★★★ Notable: The authors combine two entropy inequalities to improve the exponential upper bound for recovering pairs from (2.2499)^n to (2.23949)^n.

Multiple Access in Pinching-Antenna Systems: Survey and Impairment Evaluation

2026-10-02T00:00:00+09:00

Multiple Access in Pinching-Antenna Systems: A Survey, Impairment Analysis, and Evaluation arXiv · Oct 2, 2026This paper surveys multiple access techniques for pinching-antenna systems, which attach dielectric pinches to a waveguide to create nearby line-of-sight radiation points. It organizes fifty-six works into a unified taxonomy by spatial degrees of freedom, showing users on one waveguide can be separated only by time, frequency, power, or guided mode. The authors identify impairments that disrupt this separation and evaluate time, frequency, power, and space-domain access under identical models, finding the degradation depends on the method.

Fast demodulation for lattice-based coset codes (quantize-walk method)

2026-10-01T00:00:00+09:00

Efficient Demodulation for Coset Codes on Lattices arXiv · Oct 1, 2026This paper addresses the high complexity of soft demodulation for lattice constellations by proposing fast quantize-walk demodulation (FQWD) for multilevel coset codes. The method uses radial projections onto the Voronoi cells of the D4 and E8 lattices to reduce complexity from O(d2^{sd}) to O(sd2^d). Experiments show only negligible block error rate loss relative to maximum likelihood demodulation under AWGN.

Mechanism design for choosing a bridge location when agents have optional facility preferences

2026-09-30T00:00:00+09:00

Mechanism Design for Bridge Location with Optional Preferences arXiv · Sep 30, 2026The paper studies a bridge location problem in which two separated regions each have one prelocated facility, and each agent selects a nonempty subset of facilities she cares about. Costs are defined as the maximum, sum, or minimum distance to those facilities. The authors design deterministic strategyproof mechanisms that choose the bridge location and approximately minimize either social cost or maximum cost. They report optimal mechanisms and approximation ratios for several variants, along with lower bounds.

Non-isomorphic graphs have distinct vertex-Ramsey classes

2026-10-03T00:00:00+09:00

Non-isomorphic graphs have distinct vertex-Ramsey classes arXiv · Oct 3, 2026This paper studies the k-colour vertex Ramsey class of a graph H, the set of graphs G where any k-colouring of vertices yields a monochromatic induced copy of H. The authors prove that, for any k, non-isomorphic graphs have distinct Ramsey classes.

Single-sample prophet inequalities via reduction from combinatorial to single-item

2026-10-03T00:00:00+09:00

Single-Sample Prophet Inequalities: A Combinatorial to Single-Item Reduction arXiv · Oct 3, 2026This paper studies single-sample prophet inequalities for online combinatorial allocation, where items arrive sequentially and must be allocated. The authors give a general reduction that converts combinatorial problems into single-item prophet inequalities, using a free-disposal value to separate buyer and item constraints. The framework yields competitive ratios for XOS valuations in the single-sample and k-sample settings, and extends to divisible resources with capped-XOS valuations.

OpenMP Meta-Lowering: A Declarative Way to Generate Parallel Code Portably

2026-10-03T00:00:00+09:00

OpenMP Meta-Lowering: A Declarative Approach to Performance Portable Parallel Code Generation arXiv · Oct 3, 2026This paper presents a modular approach to parallel code generation built on MLIR, in which a domain-specific language specifies how OpenMP constructs are lowered for each target. The authors call this OpenMP meta-lowering. On PolyBench/C-OMP, it matches state-of-the-art toolchains with under 0.7% code-size overhead. The OpenMP subset is expressed in about 32% fewer lines than Clang's lowering code and about 76% fewer than GCC's, and new runtimes need only a few specification lines.

Code distances of matrix rank-metric codes under Add-and-Remove transformations: how subcode distances change when a code is modified by removing and adding subspaces.

2026-10-03T00:00:00+09:00

Code distances of matrix rank-metric codes under Add-and-Remove transformations arXiv · Oct 3, 2026The paper studies how subcode distances of matrix rank-metric codes change under the Add-and-Remove construction, which keeps a subcode of codimension ℓs and adds a space of dimension ℓa. It gives inequalities between the subcode distances of the original and new codes for arbitrary matrix codes, and shows Delsarte duality interchanges ℓs and ℓa. For matrix Gabidulin codes it gives explicit behavior of subcode distances and expresses the new code's minimum distance via rank distance between the added space and the common subcode, with relevance to the MIRANDA signature scheme.

Training one LLM across three supercomputers in two continents

2026-10-03T00:00:00+09:00

Cross-Facility LLM Pre-training on HPC: Elastic Aggregation, Data Leasing, and Queue-Aware Placement arXiv · Oct 3, 2026This paper presents a system that pools compute allocations from three supercomputers (Snellius, LUMI, and Frontier), located up to 7,400 km apart, to pre-train a single language model. It combines elastic two-loop training, a heartbeat-backed data-leasing protocol that prevents samples from being duplicated or lost, and queue-aware job placement. Training Qwen3-0.6B on C4, the three-site run reached a held-out perplexity of 34.7, compared with 28.2 for a centralised baseline. ★★★★ Notable: Combines elastic cross-site training, data leasing, and queue-aware placement; the authors report training Qwen3-0.6B across three supercomputers, reaching perplexity 34.7 versus 28.2 for a centralise

RailWave: adaptive scheduling that speeds up data transfers between expert GPUs in large AI models

2026-10-03T00:00:00+09:00

RailWave: Adaptive Spatial and Temporal Scheduling for Expert-Parallel Communication arXiv · Oct 3, 2026This paper addresses slow, uneven data exchange (All-to-All communication) in Mixture-of-Experts AI models. It presents RailWave, a communication layer built on DeepEP that redistributes traffic across network Rails and limits concurrent senders per receiver. On communication workloads from the 106B GLM-4.5-Air model, it reports speedups of up to 5.84x on H800 and 4.36x on H20 compared with Native.

Error-correcting codes with high girth built from higher-degree permutation polynomials

2026-10-02T00:00:00+09:00

Low-Density Parity-Check Codes of High Girth from Permutation Polynomials arXiv · Oct 2, 2026This paper studies quasi-cyclic LDPC codes, whose girth and minimum distance are limited by the commutativity of their circulant permutation matrices. The authors build protograph-based codes from higher-degree permutation polynomials, focusing on a family of non-commuting binomial permutation polynomials. Simulations show better decoding performance than constructions based on affine permutation polynomials.

Randomization Outperforms Deterministic Adaptive Parallel Sampling

2026-10-02T00:00:00+09:00

Randomization Beyond Deterministic Adaptivity in Parallel Sampling arXiv · Oct 2, 2026The paper studies parallel samplers that reveal each coordinate once and draw each batch independently from exact conditional marginals, under a three-round cap. For an explicit binary family, it shows randomization beats every deterministic adaptive policy: the best deterministic forward KL divergence grows linearly with dimension, while a uniform mixture of four fixed schedules is exponentially smaller. Selecting the schedule needs only two fair random bits. A finite-sum certificate confirms this separation already at dimension 389.

VenusRL: A Fully Disaggregated Agentic RL System with Priority Scheduling

2026-10-02T00:00:00+09:00

VenusRL: A Fully Disaggregated Agentic RL System with Priority Scheduling and Scalable Interaction arXiv · Oct 2, 2026This paper presents VenusRL, a fully disaggregated reinforcement learning system for training LLM agents through multi-turn tool interactions. It targets two bottlenecks: training throughput limited by the slowest trajectories, and tool sandboxes that are over-provisioned in memory. It uses priority-aware scheduling and a memory-aware page-sharing pool to address them. According to the paper, it achieves up to 4.24x end-to-end training speedup over baselines and cuts environment cost by up to 89%.

Analysis of how well a multi-rate version of NOMA (a radio access method that shares one channel among users by power) keeps data errors low when there are many users and different symbol speeds

2026-10-02T00:00:00+09:00

Performance Analysis of MR-NOMA with Arbitrary Number of Users and Symbol Rates arXiv · Oct 2, 2026This paper analytically studies the bit error rate of downlink multi-rate NOMA over Nakagami-m fading channels for any number of users and symbol rates. It derives an exact error expression for the first detected user and closed-form approximations for the others, assuming perfect successive interference cancellation. A symbol-pattern-dependent SINR analysis explains why MR-NOMA outperforms conventional single-rate NOMA, and an average BER minimization for power allocation is solved numerically and checked by simulation.

How prompt wording shapes LLMs' tendency to follow default options in group-decision games

2026-10-02T00:00:00+09:00

Prompt framing governs LLM default following in collective-action arXiv · Oct 2, 2026This paper tests whether large language models follow default values pre-filled in an interface when making choices in two one-shot social dilemmas: a common-pool resource extraction game and a threshold public-good game. Pre-filled defaults pull choices toward the default value, but the strength of this pull varies widely with wording, default type, and action-space granularity. The authors conclude that evaluating agent behaviour without controlling the choice architecture can miss a key source of variation.

AFORE: Reconfiguring Experts Efficiently in Attention-FFN Disaggregated MoE Serving

2026-10-02T00:00:00+09:00

AFORE: Attention-FFN Disaggregation with Overlapped Reconfiguration of Experts arXiv · Oct 2, 2026This paper addresses efficient serving of Mixture-of-Experts models, where uneven expert loads can slow the feed-forward stage in attention-FFN disaggregation. The authors present AFORE, which uses the upcoming microbatch expert distribution to plan expert migrations and overlaps those migrations with computation of earlier microbatches. Evaluated on a 110B-parameter MoE model over four dynamic workloads, AFORE improves output throughput by 10.1-17.6% and reduces P95 inter-token latency by 7.1-9.5% compared with the strongest baseline.

Closing the Prediction Gap: Completing Machine Shape to Match Real Hardware Behavior

2026-10-02T00:00:00+09:00

Closing the Prediction Gap: Completing Machine Shape So That Predicted Time, Power, Energy, and Mapping Match What Real Hardware Does arXiv · Oct 2, 2026This paper completes a device-shape description used to partition work across heterogeneous, multi-institution hardware ensembles without communication. It fills in five previously unmeasured fields, including memory capacity, power draw, execution unit capability, compiler directive configuration, and memory reservation overhead, and adds a Boolean reached(d) to explain an NVIDIA fp16 power shift. The fields are measured on five devices (A100, H100, V100, MI100, Max 1550), with reservation overhead ranging from about 0.004 to 0.030 and a residual of negative 9.8 percent for the excluded Triton autotuner.

PaNGEA: A GPU-Based Parallel Method for Finding Good Solutions in Mixed-Integer Optimization

2026-10-02T00:00:00+09:00

PaNGEA: Parallel Node Generation and Exploration Algorithm on GPU arXiv · Oct 2, 2026This paper introduces PaNGEA, a primal heuristic designed to run efficiently on GPUs for mixed-integer optimization problems. It combines linear-relaxation solves with a batched local-search procedure and generates multiple restricted subproblems in parallel. On 283 MIPcc26 and MIPLIB instances, the single-node GPU version reduces the average gap integral by 8-18% compared with its CPU counterpart, and parallel exploration of multiple nodes reduces it by a further 12-19%.

Coda: Using Admission Flexibility to Speed Up Coding-Agent Serving

2026-10-02T00:00:00+09:00

Coda: Exploiting Admission Flexibility for Coding-Agent Serving arXiv · Oct 2, 2026Coding agents alternate between LLM inference and tool calls, producing long sessions with reusable KV states. The paper finds two inefficiencies in shared serving systems: KV states sit across storage tiers with unequal preparation costs, and requests with different context lengths decode poorly together. It presents Coda, which uses Tiered-Aging state admission and Compatibility-Aware execution admission, plus a routing layer for multi-worker setups. Coda improves output-token and SLO-compliant throughput by 20.3% and 70.5% on average over prior systems and vLLM.

Coverage Depth in Distributed DNA Storage: When Sampling Cuts Sequencing Cost

2026-10-02T00:00:00+09:00

The Coverage Depth Problem in Distributed DNA Data Storage arXiv · Oct 2, 2026This paper studies how many DNA reads are needed to recover a full message in distributed DNA storage when strands are randomly sampled from containers. It derives exact formulas for recovery time for arbitrary linear codes and ordered partitions. The authors show MDS codes are optimal for each fixed partition and identify when distributed sampling reduces total reads rather than only parallelizing them. ★★★★ Notable: The authors give exact recovery-time formulas for distributed DNA storage, prove MDS codes optimal per partition, and resolve Bertuzzo et al.'s simplex-code conjecture.

Pricing Time, Not Just Tokens: Latency-Aware Pricing Mechanisms for LLM Inference

2026-10-01T00:00:00+09:00

Pricing Time, Not Just Tokens: Latency-Aware Mechanism Design for LLM Inference arXiv · Oct 1, 2026The paper models LLM inference as a service market in which buyers have private willingness-to-pay, task volume, and time preference, and utility depends on both token counts and latency slack. Its main result is a separation theorem: discrete hardware tiers make buyers self-select on time preference, reducing three-dimensional screening to one-dimensional screening within each tier. The authors characterize optimal tiered mechanisms using virtual-value techniques and calibrate them to 8-GPU H100 and B200 clusters.

PaxosLease in Relativistic Inertial Frames

2026-10-02T00:00:00+09:00

PaxosLease in Relativistic Inertial Frames arXiv · Oct 2, 2026The paper examines whether the PaxosLease distributed lease protocol remains safe under special relativity, where simultaneity is frame-dependent. It restates the safety property using light cones, requiring a lease expiry to lie in the causal past of any later acquisition by a different holder. The authors modify classical PaxosLease by adding a relativistic Doppler factor to its containment and quarantine rules.

ByteSplat: Faster distributed 3D Gaussian Splatting training by cutting GPU data movement

2026-10-02T00:00:00+09:00

ByteSplat: Efficient Distributed 3D Gaussian Splatting Training via Intra- and Inter-GPU communication reduction arXiv · Oct 2, 20263D Gaussian Splatting (3DGS) reconstructs photorealistic scenes, but training large scenes demands much memory and computation. ByteSplat is a distributed training framework that targets two bottlenecks: repeated off-chip memory access during rasterization and inter-GPU communication of partial gradients. It fuses forward and backward rasterization into one kernel, adds hardware-aware pruning, and removes zero gradients from communication. On six datasets, it cuts intra-GPU off-chip traffic by 63.4% and backward inter-GPU communication by 65.8%, reaching up to 6.1x training speedup on eight GPUs while preserving reconstruction quality.

Whether OFDM is optimal for aperiodic ranging signals

2026-10-02T00:00:00+09:00

On the Optimality of OFDM for Aperiodic Ranging arXiv · Oct 2, 2026This paper studies the expected aperiodic autocorrelation sidelobes of random communication symbols sent through a unitary modulation matrix, for constellations such as uniform QAM and PSK. For each lag it determines the exact minimum expected sidelobe level and a unitary modulation that attains it. The full OFDM modulation attains this minimum if and only if the lag divides the block length N, and simultaneous optimality over all positive lags is shown to be impossible for N ≥ 3. ★★★★★ Notable: The authors determine exact minimum expected sidelobe levels per lag for proper sub-Gaussian constellations such as QAM and PSK, show OFDM attains them when the lag divides the block length, and repor

AI-assisted GPU optimization of a variational method for few-body boson systems

2026-10-02T00:00:00+09:00

AI-Assisted GPU optimization of the Stochastic Variational Method for Few-Body Boson System arXiv · Oct 2, 2026The stochastic variational method is a precise approach for quantum few-body systems in fields such as nuclear and atomic physics. The authors built a parallel code for helium-4 few-body clusters, optimized for NVIDIA GH200 and AMD MI300A GPUs, with the whole code written by an AI coding agent. They introduce a secular-equation method with the Gu–Eisenstat prescription for the generalized eigenvalue problem, and tune performance by batching small matrices and optimizing memory layout. For a five-body system with 4800 basis states, the tuned code is reported to run 16.2 and 14.7 times faster than the tuned CPU version, and 168 and 153 times faster than the first CPU implementation.

ServeTwin: A Benchmark-Validated Simulator for Exploring Distributed LLM Serving Designs

2026-10-02T00:00:00+09:00

ServeTwin: A Benchmark-Validated Simulator for Distributed LLM Architecture Exploration arXiv · Oct 2, 2026This paper presents ServeTwin, a closed-loop simulator for evaluating distributed LLM serving designs without costly physical clusters. It couples specification-driven analytical timing with a stateful serving loop and can run unmodified serving benchmarks. Compared with real deployments, it reproduces a steady-state throughput-interactivity frontier with 3.6% mean error and predicts multi-turn performance with 9.9% error.

Annotation-driven migration of CUDA programs to Tenstorrent Blackhole hardware

2026-10-02T00:00:00+09:00

Annotation-Driven Migration of CUDA Programs to Tenstorrent Blackhole arXiv · Oct 2, 2026The paper presents an MLIR-based compiler that migrates statically shaped affine CUDA programs to the Tenstorrent Blackhole platform, deriving data placement, inter-core communication, and tile computation. Declarative annotations let developers specify choices the source leaves open, such as where compute and data-movement operations are placed and streaming granularity. On Gaussian elimination, a five-point stencil, and symmetric rank-k update, the best configurations reach a 4.2x speedup for BF16 Gaussian and 2.1 to 2.2x for FP32 Gaussian and Jacobi, relative to the compiler's defaults. The authors note that a choice's performance effect can reverse depending on surrounding policies, precision, and loop schedule, so alternative per-core realizations should be compared rather than policies selected independently.

Beaver: Elastic GPU Sharing for ML and Latency-Critical vRAN Workloads

2026-10-02T00:00:00+09:00

Beaver: Elastic GPU Sharing between ML and Latency-Critical vRAN Workloads arXiv · Oct 2, 2026This paper presents Beaver, a system that lets a latency-critical radio network workload and a throughput-oriented machine learning workload share the same GPU. It adjusts streaming multiprocessor allocation per time slot and rewrites ML kernels to yield memory bandwidth during the radio workload's memory-intensive phases. On an H200 GPU, the authors report that the radio workload's p99.9 latency stays within its 1.5ms uplink deadline while retaining 74% of Llama-3.3-70B serving throughput.

Existence of Two-Sided Product Expanding Codes Using Rademacher Matrices

2026-10-02T00:00:00+09:00

Two-Sided Product Expanding Codes via Rademacher Matrices arXiv · Oct 2, 2026The paper proves that two-sided product expanding codes exist without relying on explicit constructions of asymptotically optimal locally testable codes (c³-LTCs), unlike an earlier result by Kalachev and Panteleev. It shows that independent random linear codes over a sufficiently large prime field have constant two-sided product expansion with probability tending to one, for fixed numbers of component codes with rates bounded away from zero and one. The trade-off is that the field characteristic must grow with the block length, whereas the earlier work used extensions of F₂. ★★★★ Notable: The authors replace explicit c^3-LTC constructions with random Rademacher matrices over the reals and use the trace power method to show two-sided product expanding codes exist; they report random lin

Burning Signed Graphs: Two competing fires spread on signed graphs, aiming to burn as many vertices of one colour as possible.

2026-10-02T00:00:00+09:00

Burning Signed Graphs arXiv · Oct 2, 2026The paper introduces a new graph burning model in which yellow and green fires ignite vertices of a signed graph and spread along edges, with an edge's sign determining whether the fire changes colour. Each step, a player ignites a new vertex in a chosen colour. The optimal achievable count, called the plurality number, is expressed through Hamming distances to a binary code associated with the graph. The authors determine exact plurality numbers for certain signed paths and prove hardness results for computing or approximating it.

Optimizing simulation and data pipelines for experiments on exascale computers

2026-10-02T00:00:00+09:00

Real-time Optimization of Simulation and Data Processing Pipelines for Experiments on Exascale Computing Platforms arXiv · Oct 2, 2026This paper addresses how large experiments in high-energy physics and cosmology, which generate growing volumes of data, can use exascale computing resources efficiently. The authors propose a multi-stage task scheduling approach and demonstrate it on a Monte Carlo simulation pipeline from the Short-Baseline Near Detector (SBND). They also present theoretical and numerical tools for modeling task scheduling, which can help identify optimal strategies and choose resources for a given problem size.

Conditional Correctness in List Decoding: When a Later Second Codeword Boosts Confidence in the First

2026-10-02T00:00:00+09:00

Conditional Correctness in List Decoding: How a Late Second Codeword Can Rescue Confidence in the First arXiv · Oct 2, 2026This paper studies how the likelihood ranks of noise sequences in list decoding relate to the posterior probability that a decoding is correct. Under a random codebook model, the authors show that soft-output expressions for list decoding exactly match finite-blocklength conditional correctness probabilities, and they characterize the asymptotic reliability of the first list entry under hard-decision maximum likelihood decoding. The reliability depends only on the first, second, and last ranks. For list size two, a sufficiently late second entry can make the first entry asymptotically correct even when its rank alone suggests otherwise.

Topology-aware integrated sensing, communication, and charging for large-scale low-altitude wireless networks

2026-10-02T00:00:00+09:00

Topology-Aware Integrated Sensing, Communication, Charging in Massive Low-Altitude Wireless Network arXiv · Oct 2, 2026This paper addresses low-altitude wireless networks that must support sensing, communication, and charging at once, which creates tightly coupled optimization problems. The authors propose a unified topology-aware framework that represents terrestrial and airborne devices as nodes and their interactions as edges in a bipartite graph. Based on this representation, they develop a low-complexity topology reconfiguration algorithm using a unified resource-adjustment rule. Simulations show better multi-functional coordination efficiency than state-of-the-art methods, with robustness and scalability in scenarios with many ground and airborne users.

From Alert Floods to Precedence Forests: LOGOS, a Zero-Prior-Knowledge Incident Triage System

2026-10-02T00:00:00+09:00

From Alert Floods to Precedence Forests: Zero-Prior-Knowledge Incident Triage with LOGOS arXiv · Oct 2, 2026This paper asks whether failure propagation can be isolated using only raw, unannotated text logs. The authors present LOGOS, an unsupervised system that uses entity-event co-occurrence and temporal precedence to compress millions of log lines into a compact precedence forest. Tested on 25 production enterprise outages and 12 open-source issues without any prior knowledge, it reports a median runtime of 4.5 minutes, a median 99.8% noise reduction, and a mean recall of 0.76.

Feature freshness budgets for real-time ML inference under stream lag

2026-10-01T00:00:00+09:00

Feature Freshness Budgets for Real-Time ML Inference Under Stream Lag arXiv · Oct 1, 2026The paper models an online feature store as a materialized view over an event stream and defines a per-feature freshness budget, the gap between a feature's decision window and the staleness imposed by serving. It proves a per-feature staleness bound and a closed-form threshold below which a feature cannot be served within budget. The results are validated in a discrete-event simulation and on a real Kafka, Redis, and PostgreSQL pipeline, where the predicted threshold reproduces. The authors find the main simulation-to-infrastructure gap stems from sampling-phase artifacts rather than latency, and randomizing the phase reduces the drop-rate error to 0.018. ★★★★ Notable: Defines online feature freshness as a bounded budget with a closed-form threshold, validated on real Kafka, Redis, and PostgreSQL pipelines; the authors report that phase-locked simulation systematica

MoE-CORE: A system that coordinates expert offloading and storage to run large Mixture-of-Experts models on limited device memory.

2026-10-02T00:00:00+09:00

MoE-CORE: Coordinated Expert Offloading and Residency for Memory-Constrained MoE Inference arXiv · Oct 2, 2026This paper presents MoE-CORE, a system for running Mixture-of-Experts (MoE) model inference when device memory is limited. It loads complete expert layers in alternating buffers during prefill and, during decoding, uses layer-wise cache sizing, routing-history-aware replacement, and cross-layer prefetching to reduce host-to-device transfers. The authors report substantially lower mean time per output token than the evaluated vLLM Prefetch configuration on DeepSeek-V4-Flash-W4A8 and GLM-5.2-W4A8C8.

ePACT: An energy-performance-aware controller that tracks hourly power commitments for LLM serving

2026-10-01T00:00:00+09:00

ePACT: Energy-Performance-Aware Commitment Tracking for LLM Serving arXiv · Oct 1, 2026The paper argues that cutting LLM serving energy alone does not guarantee lower costs, because deviations from preset electricity commitments can be costly, especially when overuse and underuse are priced asymmetrically. The authors propose ePACT, a two-level controller that adjusts serving capacity and GPU clocks while meeting request-level service requirements. In full-day trace simulations on H20 and H200 GPU pools, ePACT reduces the asymmetric deviation cost by 73.8% and 75.7% relative to vLLM while retaining near-vLLM SLO attainment.

Storage codes from binary BCH codes that approach unit rate

2026-10-01T00:00:00+09:00

Asymptotically unit-rate storage codes from binary BCH codes arXiv · Oct 1, 2026A storage code assigns symbols to graph vertices so each symbol can be recovered from its neighbors. The paper proves that binary full-parity storage codes on triangle-free coset graphs of primitive BCH codes have rate approaching one for any fixed error-correction capability of at least two. For the double-error-correcting family over F_{2^m}, the storage deficiency is shown to be Θ(((1+√5)/4)^m).

Token-based economy for fairer, efficient highway congestion control using express lanes

2026-10-01T00:00:00+09:00

Token Economy Design for Fair and Efficient Highway Congestion Management with Express Lanes arXiv · Oct 1, 2026This paper studies a non-monetary token economy for allocating highway lanes, aiming to improve fairness without reducing traffic efficiency. High-occupancy vehicles get unrestricted express lane access, while other users earn and spend tokens to switch between regular and express lanes. In a microscopic traffic simulation using real-world demand data, the proposed prices produced nearly the same average travel time as a scenario with no reserved express lane, while substantially reducing urgency-weighted perceived travel time.

Giving LLMs social value profiles makes their play in a social dilemma more human-like

2026-10-01T00:00:00+09:00

Conditioning LLMs on Social Value Orientation improves behavioural alignment in a sequential social dilemma arXiv · Oct 1, 2026The authors test eight large language models on two versions of the Centipede Game and compare their default choices with human experimental data. The models' default strategies vary widely and are generally far from the human reference. When conditioned on Social Value Orientation (SVO) profiles drawn from human participants, alignment with human behaviour improves by up to 70%, and higher SVO values reduce the probability of stopping the game.

Speeding up MPI programs by letting compute-heavy and memory-heavy parts drift out of step

2026-10-01T00:00:00+09:00

Exploiting the Interplay of Compute- and Memory-Bound kernels in MPI Applications arXiv · Oct 1, 2026The paper shows that in MPI applications with little communication that alternate between compute-bound and memory-bound phases, communication stalls can unintentionally ease memory-bandwidth contention. The authors demonstrate this with a parallel optical flow solver combining ray tracing and an optical flow kernel, achieving considerable speedup through desynchronization. A tunable microbenchmark and a bandwidth-aware simulator are used to study and validate the effect.

Optimal Universal Coding of Integers: Numerically Determining the Minimum Expansion Factor

2026-10-01T00:00:00+09:00

Optimal Universal Coding of Integers arXiv · Oct 1, 2026This paper studies universal coding of integers (UCI), which assigns binary codewords to positive integers so that the average codeword length stays within K times max{1,H(P)} for every nonincreasing source distribution. The authors show that the worst-case ratio for any universal code is attained by a family of one-point-plus-uniform-tail distributions. They also introduce a UCI inequality, analogous to the Kraft inequality for prefix codes, which yields an equivalent definition of the optimal minimum expansion factor C*. Numerical computation gives C* = 2.000124757036101…, with the first fifteen decimal digits certified.

Agentic Wireless Digital Twin Building and Calibration from Real Measurements

2026-10-01T00:00:00+09:00

Agentic Wireless Digital Twin Construction and Calibration Using Real-World Measurements arXiv · Oct 1, 2026This paper proposes AWDT, an agent-based framework that automatically builds and calibrates wireless digital twins. Three agents handle the propagation environment, base station and sector configuration, and radio material properties, refining the model using differences between smartphone-measured RSRP and ray-tracing predictions. In real-world experiments, the RSRP prediction MAE dropped from 11.46 dB to 5.25 dB. Evaluation with an independent measurement system also showed transfer across devices with lightweight device-specific bias adaptation.

GPU-Initiated Communication: A Low-Level Performance Dissection

2026-10-01T00:00:00+09:00

GPU-Initiated Communication: Dissecting Down to the Bone arXiv · Oct 1, 2026This paper examines how GPU threads post RDMA operations directly to the network interface card, a mechanism used by NVSHMEM, NCCL GIN, and DeepEP. The authors build minimal GPU and CPU-proxy transports and benchmark them against several libraries on NVIDIA H100, H200, B200, and GB200 systems. They report that a minimal GPU path issues an operation in 0.7 µs and completes it in 4.0 µs, with library overheads adding up to 4.6 µs. The paper concludes that the submission path alone does not predict overall communication performance. ★★★★ Notable: Dissects the GPU-to-NIC path with minimal transports mini-gda and mini-proxy, comparing against several libraries on H100, H200, B200 and GB200, and separating hardware costs from library overhead.

Joint communication and sensing coverage for UAV corridors with beam misalignment

2026-10-01T00:00:00+09:00

Communication and Sensing Coverage of UAV Corridor under Beam Misalignment Error arXiv · Oct 1, 2026This paper studies joint communication and sensing (JCAS) coverage probability for a terrestrial base station using a realistic 3D 3GPP antenna pattern, with UAV users deployed in a finite cylindrical aerial corridor. The base station uses radar sensing to track a known UAV, under a 3D beam misalignment error modeled as multivariate normal. Exact analytical expressions for JCAS coverage probability are derived under clutter and uplink interference. The results show corridor-length saturation at high UAV heights and a beamwidth trade-off between interference suppression and misalignment sensitivity.

Joint Communication and Sensing in Aerial Corridors: A Stochastic Geometry Model

2026-10-01T00:00:00+09:00

Joint Communication and Sensing in Aerial Corridors: A Novel Stochastic Geometry Framework arXiv · Oct 1, 2026This paper studies how well a drone base station can both communicate and sense in an aerial corridor filled with drone users. The corridor is modeled as a finite cylinder, with users randomly placed in three dimensions and the base station tracking a known user with radar. The authors derive exact formulas for coverage probability under clutter and uplink interference. They find that more directional base station antenna beams improve performance noticeably, especially in shorter corridors.

UAV attitude-aware trajectory and attitude co-design for covert communication

2026-10-01T00:00:00+09:00

Exploiting UAV Attitude for Covert Communications: Dynamics-Consistent Trajectory-Attitude Co-Design arXiv · Oct 1, 2026Existing UAV covert communication trajectory designs usually treat the drone as a point mass and ignore how maneuvers affect antenna orientation. This paper studies a rotary-wing UAV with a directional antenna sending confidential data to a legitimate receiver while a warden is present, explicitly modeling acceleration-induced attitude and its effect on channel gain. It derives a covertness constraint using KL divergence at the warden and jointly optimizes trajectory and attitude with an MM-based algorithm to maximize average covert rate. The framework is also extended to a robust design under imperfect attitude control.

RapidMoE: Adaptive Residual Offloading for Faster Large-Scale MoE Inference

2026-10-01T00:00:00+09:00

RapidMoE: Exploiting Cross-Asymmetry via Adaptive Residual Offloading for Large-Scale MoE Inference arXiv · Oct 1, 2026This paper addresses the mismatch between large Mixture-of-Experts (MoE) models and heterogeneous CPU-GPU hardware, where PCIe bandwidth limits and heavy CPU computation reduce efficiency. It proposes RapidMoE, a residual offloading system that shifts offloading from the expert level to the bit level, using compact data representation, dual-path routing matched to hardware, and balanced scheduling. The authors also introduce a runtime arbitration method to adjust critical experts. Experiments report up to 3.5x decoding speedup and 2.1x prefill speedup over state-of-the-art offloading systems.

Serving Revisable Agent Tasks Through Versioned Execution

2026-10-01T00:00:00+09:00

Serving a Revisable World: Versioned Execution for Interruptible Agents arXiv · Oct 1, 2026The paper presents RETIRE, a serving control-plane design for LLM agents whose running tasks get revised. Each execution version holds the authority to publish output or install state, so obsolete work is revoked first and the completed prefix that a successor can inherit is certified. The design is implemented in vLLM. In controlled paired experiments, it reduces the median revision-to-successor time-to-first-token by 17.1%. ★★★★ Notable: Proposes versioned execution to unify revoking obsolete work with inheriting completed KV state, implemented in vLLM; the authors report a 17.1% median TTFT reduction in controlled paired experiments.

Fair, non-wasteful matching of families and daycare places, including siblings and already-enrolled children

2026-10-01T00:00:00+09:00

Fairness and Non-wastefulness in Matching with Siblings and Initial Enrollments: Compatibility and Complexity arXiv · Oct 1, 2026This paper studies matching between families and daycare centers, where a family may have several children with joint preferences and some children are already enrolled and seeking transfers. The authors introduce resettlement, which lets displaced families return to their initial assignments, and give a greedy algorithm showing that matchings satisfying non-wastefulness and resettlement-based fairness exist. They also analyze computational complexity, finding that some existence questions are NP-complete and computing resettlement-based outcomes is PLS-hard.

Seamless reconfiguration for DAG-based Byzantine atomic broadcast in asynchronous networks

2026-10-01T00:00:00+09:00

Seamless Reconfiguration for DAG-BFT arXiv · Oct 1, 2026This paper proposes a seamless reconfiguration protocol for asynchronous DAG-based Byzantine atomic broadcast, which adds or removes nodes. Existing DAG-based BFT protocols assume a static set of processes, and prior reconfiguration approaches either pause the system or wait for a predetermined epoch boundary. The proposed protocol lets the DAG keep growing while membership and quorum thresholds change, without halting the system. The authors state it is the first such protocol for fully asynchronous networks.

Reading While Writing: Baseline Data Needs for Molecular Neural Interfaces

2026-10-01T00:00:00+09:00

Reading While Writing: Baseline Information Requirements for Molecular Neural Interfaces arXiv · Oct 1, 2026This paper studies molecular neural interfaces that both deliver and sense a neurotransmitter, and how they can estimate endogenous release despite their own chemical input. The authors analyze two exchanging regions with shared saturable uptake and show that an unknown resting concentration difference leaves the source region's uptake uncertain. A dopamine-inspired example and a Gaussian observation model are used to show how delivery and baseline information affect release classification.

MANTA: ML-Augmented Tiering Advisor for Smarter Memory Page Placement

2026-10-01T00:00:00+09:00

MANTA: Machine Learning Augmented Tiering Advisor arXiv · Oct 1, 2026Memory tiering combines fast DRAM with slower tiers such as CXL-attached memory to expand capacity. The paper introduces ChOMP, an offline optimizer used as a reference, and a trace-driven simulator that reveals opportunities for online policies. It then presents MANTA, which predicts page usefulness from runtime access features using a lightweight learned model integrated into ARMS. On eight emulated CXL workloads at 4 GB fast memory, MANTA achieves geometric-mean speedups over ARMS of 1.12× on Linux 6.2 and 1.08× on Linux 6.18, and 1.69× across six Optane workloads.

Ditto: Generalized Reconfigurable Linearizable Reads

2026-10-01T00:00:00+09:00

Ditto: Generalized Reconfigurable Linearizable Reads arXiv · Oct 1, 2026This paper studies how to make linearizable reads faster in distributed datastores without giving up consistency. It classifies existing read algorithms into a generalized space and analyzes their latencies mathematically, revealing two fundamental tradeoffs. The authors propose a new algorithm, Pairwise Quorums, and Ditto, a system that reconfigures across this space at runtime as network and workload conditions change. They claim that together these subsume existing algorithms, being either strictly better or matching them where a real tradeoff exists.

MegaFlux: Skew-resilient MoE megakernels with pipelined expert replication

2026-10-01T00:00:00+09:00

MegaFlux: Skew-Resilient MoE Megakernels via Pipelined Expert Replication arXiv · Oct 1, 2026This paper addresses slowdowns in mixture-of-experts (MoE) models, where uneven routing overloads some GPUs while others sit idle. It presents MegaFlux, which decides at runtime whether to replicate popular experts and pipelines the resulting weight transfers and backward gradient reductions with expert computation. Across 147 configurations per direction on eight NVIDIA B200 GPUs, it reports geometric-mean speedups of 1.45x for forward and 1.28x for backward over fixed expert placement.

Forming Alliances Under Uncertainty: A General Lotto Study

2026-10-01T00:00:00+09:00

Forming Alliances in the Fog of War: A General Lotto Perspective arXiv · Oct 1, 2026This paper studies whether two players fighting a shared opponent can benefit from alliances when they only partly know each other's or the adversary's resource budgets. It uses the coalitional General Lotto game, where players allocate budgets across valued contests. The authors characterize which games admit a budget transfer that helps both players under multiplicative uncertainty. They show such robust transfers exist in a positive-measure set of games at every uncertainty level.

Shaping and modulation augmentation for LDPC-coded QPSK

2026-10-01T00:00:00+09:00

Modulation Augmentation and Constellation Shaping for LDPC-Coded QPSK arXiv · Oct 1, 2026This paper adds a short shaping code to LDPC-coded QPSK so that some codeword bits control how often lower- and higher-energy constellation points are sent. The lower-energy points are chosen more often, producing a Gaussian-like symbol distribution. The receiver refines shaping-bit likelihoods using LDPC decoder feedback, and the shaping codebook, bit mapping and constellations are jointly optimized. The authors report gains of up to 0.6 dB over 5G NR LDPC-coded QPSK and 0.53 dB over QAM-16.

Market design for two-sided ad auctions with autobidders: which mechanisms still work when both buyers and sellers are strategic

2026-10-01T00:00:00+09:00

Two-sided Market Design Meets Autobidding arXiv · Oct 1, 2026The paper studies two-sided markets in which autobidders, such as Return-on-Spend (RoS) value maximizers, participate on one or both sides. It shows that, without prior information, many incentive-compatible, budget-balanced mechanisms have unbounded Price of Anarchy for liquid gains from trade when value maximizers are present. It also finds that McAfee's Trade Reduction mechanism keeps a constant Price of Anarchy for liquid welfare under RoS bidding. With distributional information, the authors design a mechanism that achieves first-best liquid welfare when at least one side consists of RoS maximizers, contrasting with the Myerson-Satterthwaite impossibility. ★★★★ Notable: First study of two-sided autobidder markets: the authors report unbounded PoA for broad mechanism classes prior-free, constant liquid-welfare PoA for McAfee's mechanism, and a first-best mechanism und

L2R-EV: Learning Which Repairs to Make in Electric Ride-Pooling with Limited Charger Queues

2026-10-01T00:00:00+09:00

L2R-EV: Learning What to Repair in Electric Ride-Pooling with Finite Charger Queues arXiv · Oct 1, 2026The paper studies electric ride-pooling, where passenger service, routes, batteries, and limited chargers must be managed together. The authors build a discrete-event simulator and add a learned repair layer that reassigns uncollected passengers, guided by a supervised score and a two-stage PPO policy. On 12 held-out Manhattan episodes, supervised repair cut operational cost by 6.57% versus no repair and won in all 12 episodes, while queue-aware charging reduced waiting by 41.0%.

LatencyLab: a DPDK-based tool for measuring P4 pipeline latency on FPGA SmartNICs

2026-10-01T00:00:00+09:00

LatencyLab: A DPDK-Based P4 Pipeline Latency Measurement Framework for FPGA SmartNICs arXiv · Oct 1, 2026This paper presents LatencyLab, a DPDK-based framework that measures how much latency a P4 program adds on an FPGA SmartNIC, without needing PTP support or clock synchronization. A switch duplicates each probe packet to two ports, one through the VitisNetP4 pipeline and one through a matched bypass path, and the arrival-time difference isolates pipeline latency. The authors evaluated four VitisNetP4 programs on an AMD Alveo U280 and report tight, reproducible latency distributions, with two independent checks agreeing with the measurements.[18]

Generalized Prophet Inequalities for Online Auctions: Agents with Budgets and Mixed Objectives

2026-09-30T00:00:00+09:00

Generalized Prophet Inequalities for Online Auctions arXiv · Sep 30, 2026This paper studies prophet inequalities for online auctions, a setting that classically assumes agents maximize quasi-linear utility. The authors model agents' budgets and varying objectives, such as value maximization in autobidding, as stochastic inputs alongside valuations. They analyze anonymous posted-price mechanisms and generalize the balanced prices framework to incorporate budgets and mixed objectives. The paper shows constant competitive ratios in several central settings, including agents with subadditive valuations, and identifies trade-offs between competitive ratio and price sensitivity. It also gives a general reduction from budgeted settings to budget-free value maximizers. ★★★★ Notable: The authors extend prophet inequalities to budgeted agents with mixed, stochastic objectives, report constant competitive ratios for balanced-price mechanisms in several settings, and give a reduction

Darpan: A Digital Twin Framework for Next-Generation Computing Continuum Decisions

2026-09-30T00:00:00+09:00

Darpan: A Digital Twin Framework for the Next-Generation Computing Continuum arXiv · Sep 30, 2026This paper proposes Darpan, a digital twin framework for applications spanning devices, edge, fog, and cloud. A physical side runs real applications and observes their state, while a digital side keeps an executable virtual copy updated to evaluate candidate decisions from the same starting point. The selected decision is validated against the latest physical state before execution. The authors report a mean absolute error of 0.485 s for predicted response time and 93.1% scale-out efficiency at 40 nodes on real DAG workloads.

Adaptive power allocation to improve mission efficiency for coordinated low-altitude aircraft swarms

2026-09-30T00:00:00+09:00

Mission Efficiency Optimization in Low-Altitude Economy: Adaptive Power Allocation for Coordinating Heterogeneous Aircraft Swarms arXiv · Sep 30, 2026This paper studies how multiple heterogeneous low-altitude aircraft cooperate on complex missions, with some handling sensing, communication, computing, and execution. The authors introduce operational-capability entropy to quantify each aircraft's effective work capability. They formulate a power allocation problem that minimizes a linear quadratic regulator cost as a mission-efficiency metric, considering channel conditions. The problem is split into two convex subproblems solved iteratively with closed-form solutions, and simulations show the scheme outperforms traditional approaches.

Comparing adaptive Byzantine quorum systems to reduce latency in wide-area networks

2026-09-30T00:00:00+09:00

Exploring Adaptive Byzantine Quorum Systems to Improve Latency in the WAN arXiv · Sep 30, 2026In Byzantine fault-tolerant replication, a quorum of replicas must vote on a value before it is decided, and the quorum's size and makeup affect client latency over wide-area networks. Existing quorum designs are usually tied to different protocol implementations, which makes fair comparison difficult. The authors extend BFT-SMaRt with a modular quorum-system abstraction so that several quorum constructions can run in the same framework with the same optimizations. They then use network simulation to study how these constructions affect client-observed latency.

Age of information distributions in an Erlang loss system with shared servers

2026-09-30T00:00:00+09:00

Distribution of Age of Information in the Erlang Loss System arXiv · Sep 30, 2026This paper derives exact distributions of age of information (AoI) and peak AoI (PAoI) in a bufferless M/M/c/c Erlang loss system, where Poisson-arriving updates from one or more sources share c servers. Three policies handle arrivals when all servers are busy: blocking, preempting a random same-source update, or preempting the stalest one. Using an absorbing Markov chain method, the authors obtain matrix-exponential distributions, which extend to multiple sources affected only by aggregate update rates.

Kirin: An orchestrator for cloud-native WebAssembly services

2026-09-30T00:00:00+09:00

Kirin: Cloud-native WebAssembly Service Orchestration arXiv · Sep 30, 2026This paper presents Kirin, an orchestrator for cloud-native WebAssembly services that handles resource scheduling, lifecycle management, scaling, and request routing. The authors evaluate Kirin against container-based deployments across several service workloads. Results are promising for service lifecycle management, while known limitations remain for CPU-intensive workloads. For typical service workloads, Kirin performs competitively, suggesting WebAssembly could be a viable cloud-native deployment option.

Tight liquid welfare bounds for budget-constrained auctions via LP duality

2026-09-30T00:00:00+09:00

Tight Liquid Welfare Guarantees for Auctions with Budgets via LP Duality arXiv · Sep 30, 2026This paper studies the efficiency of auctions with budget-constrained bidders and derives tight price of anarchy bounds for coarse correlated equilibria. It introduces a general framework that reduces these guarantees to finding feasible solutions to the dual of a linear program, along with a new smoothness notion for budget-constrained settings. Using this framework, the authors obtain tight liquid welfare guarantees for many auction formats and new lower bounds for the budget-free uniform price and generalized second-price auctions. ★★★★★ Notable: Proposes an LP-duality framework and a new smoothness notion for budget-constrained auctions, deriving tight PoA bounds for many formats and new lower bounds for uniform price and GSP auctions; the un

Version Age-of-Information-Based Semantic Secrecy Region Analysis: examines semantic secrecy in wireless status updates, balancing information freshness against eavesdropping.

2026-09-30T00:00:00+09:00

Version Age-of-Information-based Semantic Secrecy Region Analysis arXiv · Sep 30, 2026This paper studies semantic secrecy in wireless status-update systems using Version Age of Information (VAoI). It models the tradeoff between informative updates and confidentiality with a three-node wiretap channel. The authors introduce the semantic secrecy region (SSR), analyze it geometrically, and derive closed-form expressions for the destination's average VAoI and the eavesdropper's semantic asymmetry probability using a two-dimensional Markov chain. For Rayleigh fading channels, they characterize the optimal SSR boundary and power allocation, and numerical results show that optimized power allocation enlarges the secrecy region compared with fixed-power benchmarks.

Federated Calibration of UAV Motion Uncertainty in Multi-Operator ISAC

2026-09-30T00:00:00+09:00

Federated Calibration of Motion Uncertainty for UAV Tracking in Multi-Operator ISAC Systems arXiv · Sep 30, 2026This paper studies UAV tracking in integrated sensing and communication (ISAC) systems with multiple operators, where raw sensing data, channel information, and transmit decisions stay local. The authors propose a two-timescale framework: slow-timescale calibration of motion uncertainty using local radar innovations, and online fusion of track estimates via covariance intersection with predictive beamforming. Simulations show improved tracking-uncertainty consistency under motion model mismatch, and about 31% lower RMSE than communication-only transmission design with less than 2% extra power.

Fair Outcome Prospects Encourage Prosocial Behaviour in the Centipede Game

2026-09-30T00:00:00+09:00

The prospect of a fair outcome triggers prosocial behaviour in the Centipede Game arXiv · Sep 30, 2026This behavioural experiment studies how people make sequential resource-exchange decisions in the Incremental Centipede Game, comparing a zero-payoff deadline treatment with a treatment ending in a fair split. Even though the fair outcome is uncertain, it moves participants beyond self-interest. An evolutionary model indicates that the fair split shifts play toward mutually beneficial strategies when players are sufficiently prosocial and use probabilistic strategies.

FissionReady: Joint Workload and Power Scheduling for Small Modular Reactor Data Centers

2026-09-30T00:00:00+09:00

FissionReady: Joint Workload and Power Scheduling for Data Centers Powered by Small Modular Reactors arXiv · Sep 30, 2026This paper presents FissionReady, a scheduler for data centers powered by small modular reactors (SMRs). It tracks each module's fuel age, xenon state, and remaining flexibility to set per-module operating limits, then coordinates load-following, batch deferral, and grid purchases through a two-timescale optimization. Compared with a same-sized base configuration, the authors report zero involuntary shutdowns and zero batch deadline misses, with water use reduced by about one-third and grid cost by about half.

Communication-efficient (1+ε)Δ edge coloring and the Lovász local lemma

2026-09-30T00:00:00+09:00

Communication-Efficient (1+ǎrepsilon)Δ-Edge Coloring and Lovász Local Lemma arXiv · Sep 30, 2026This paper studies edge coloring in the two-party edge-partition model, where Alice and Bob each hold part of the edges and must jointly produce a proper coloring with little communication. The authors give a public-coin Las Vegas protocol that finds a (1+ε)Δ edge coloring for sufficiently large Δ, using expected O(ne^{-γΔ}+1) bits and O(log n/Δ+1) rounds. They also develop a two-party constructive Lovász local lemma, which they apply to several other classical problems. ★★★★ Notable: Proposes a new randomized coloring procedure and a two-party constructive LLL, reducing colors to (1+ε)Δ and communication to o(n) bits; the authors also apply the LLL to other classical problems.

Multi-View LP Bounds for Locally Recoverable Codes with Availability

2026-09-30T00:00:00+09:00

Multi-View Block Distance Distributions and Linear Programming Bounds for Locally Recoverable Codes with Availability arXiv · Sep 30, 2026The paper develops a multi-view linear-programming framework for locally recoverable codes with fixed availability a, recording Hamming distances across partitioned helper-set blocks. Constraints from counting identities, product-Krawtchouk positivity, and collision inequalities give a polynomial-size relaxation for arbitrary finite-field codes. For availability 2, exact rational certificates establish M_max values of 8, 27 and 64 for three parameter sets. The four-block bound is strictly stronger than both coarser models.

Vosti: A Verified Engine for Bitwise-Deterministic LLM Inference

2026-09-30T00:00:00+09:00

Vosti: Specifying, Implementing, and Verifying Deterministic LLM Inference arXiv · Sep 30, 2026The paper formally defines deterministic LLM inference as producing bitwise-identical logits for the same prompt and initial sampler state under a fixed model and deployment configuration. The authors find that production modes such as vLLM's batch-invariant mode and SGLang's deterministic mode yield different logits under some execution variations. They present Vosti, an inference engine built against this specification, with a Verus inductive proof for scheduling and the prefix-sharing KV cache and a Triton analyzer proving kernel output equality. Vosti reports identical logits across all tested variations and performance comparable to vLLM's batch-invariant mode on decode-heavy workloads. ★★★★ Notable: Formalizes deterministic LLM inference; the abstract reports that Vosti uses a Verus inductive proof to show scheduling and the paged KV-cache preserve logits, and a Triton analyzer to prove selected

Exact embeddings of complete bipartite graphs and non-embeddable lattices under information distance

2026-09-30T00:00:00+09:00

Complete bipartite embeddings, lattice non-embeddability, and non-Euclidean powers of information distance arXiv · Sep 30, 2026The paper encodes vertices of finite complete bipartite graphs as binary strings at each scale, so that information distances approximate scaled graph distances with a constant additive error. This resolves an open problem of Hutter on scale-embeddings of such graphs and shows that no positive power of information distance is exactly representable in a real Hilbert space. The authors also prove that integer lattices admit no scale-embedding and extend this conclusion to (R^m, ||·||_1). ★★★★ Notable: The authors give string embeddings of complete bipartite graphs and use them to answer two of Hutter's questions on information distance powers and lattice scale-embeddings.

Efficient HPC Systems for AI Agents: Challenges and Opportunities

2026-09-30T00:00:00+09:00

Towards Efficient HPC Systems for Agents: Challenges and Opportunities arXiv · Sep 30, 2026This paper examines how coding agents have become real users of high-performance computing systems, while HPC abstractions, interfaces, and policies were designed for human-driven workflows. The authors report that agent users are 19.5% of observed users but account for 55.8% of job submissions, 29.1% of CPU core-hours, and 42.7% of GPU-hours. They argue that facilities should co-design systems that treat agents as first-class principals, and outline challenges and opportunities in compute, storage, agent memory, and safety.

Keeping Provenance in Shared KV Caches to Prevent Unsafe Reuse in LLM Serving

2026-09-30T00:00:00+09:00

Preserving Provenance in Shared KV Caches for LLM Serving arXiv · Sep 30, 2026The paper finds that LLM serving systems keep provenance such as adapter and weight configuration in local prefix caches, but the shared KV-cache tier keys entries only by token content and coarse model metadata. This lets identical tokens from incompatible contexts collide; cross-adapter collisions cut accuracy from 0.94 to 0.64 in the authors' experiments on vLLM and SGLang. The authors propose a KV provenance contract that binds provenance into lookup and store keys through a canonical descriptor, implemented across three cache paths. The implementation eliminates unsafe reuse while keeping legitimate sharing, with hit-path latency changes within 0.34 ms. ★★★★ Notable: The paper gives the first systematic study of provenance loss in shared KV caches and proposes an injectable provenance contract implemented in vLLM and SGLang; the authors report cross-adapter collis

Efficiency of Generalized Proportional First-Price Auctions Under Auto-bidding

2026-09-30T00:00:00+09:00

Efficiency of Generalized Proportional First-Price Auctions Under Auto-bidding arXiv · Sep 30, 2026This paper studies how efficiently auctions allocate ad slots when advertisers use auto-bidding with return-on-spend constraints. It measures efficiency loss via the Price of Anarchy across prior-free auction formats. For two bidders, the standard proportional first-price auction achieves a tight PoA of at most 1.5. For n bidders, setting the exponent to 2n yields a PoA of 2 minus 1/(4n+1), beating the deterministic barrier of 2. ★★★★ Notable: Studies efficiency loss of generalized proportional first-price auctions under auto-bidding. The authors report a PoA upper bound of 1.5 for two bidders with the standard proportional first-price auct

Efficient client selection for federated learning as a service in multi-application IoT systems

2026-09-30T00:00:00+09:00

Enabling Efficient Client Selection in FL-as-a-Service for Multi-Application based Society 5.0 arXiv · Sep 30, 2026This paper examines federated learning as a service (FLaaS) in IoT, where models are trained across devices without sharing raw data, enabling collaboration across applications. The authors propose selecting client nodes for model training based on delay, energy consumption, and link status. They evaluate the method on a Human Activity Recognition dataset under simulated IoT network conditions, comparing it with random and Q-learning based selection. The reported results show improved delay and energy efficiency while maintaining communication performance.

Auctions with Price Predictions: Designing Single-Item Auctions Using a Predicted Revenue-Maximizing Price

2026-09-30T00:00:00+09:00

Auctions with Price Predictions arXiv · Sep 30, 2026The paper designs auctions for selling one item in unlimited supply, given only a single prediction of the revenue-maximizing uniform price rather than predictions of each bidder's value. The authors characterize the Pareto frontier of consistency and robustness achievable by any universally truthful auction. They show that randomizing between posting the predicted price and running an optimal prior-free fallback auction is Pareto-optimal. The paper also extends the results to graceful revenue degradation with prediction accuracy and studies learning such predictions from historical market data. ★★★★ Notable: Designs single-item auctions from one predicted revenue-maximizing price, characterizes the consistency-robustness Pareto frontier, and gives a learning path from historical data.

Rotatable antennas for space-air-ground networks: opportunities and challenges

2026-09-29T00:00:00+09:00

Rotatable Antenna-Enabled Space-Air-Ground Integrated Networks: Opportunities and Challenges arXiv · Sep 29, 2026This article examines how rotatable antenna (RA) technology, which adjusts beam direction mechanically or electronically while keeping antenna positions fixed, can support communication and sensing in space-air-ground integrated networks (SAGINs). It explains RA's role in wide-area coverage, dynamic directional transmission, and cooperative sensing, and discusses design challenges such as channel acquisition and cross-segment coordination. An RA array prototype and two simulation studies are presented to illustrate feasibility and potential performance gains.

Why does scale-invariant regret matching converge linearly at the last iterate?

2026-09-30T00:00:00+09:00

Why Last-Iterate Scale-Invariant Regret Matching Converges Linearly? arXiv · Sep 30, 2026The paper studies IREG-PRM+, which normalizes cumulative regret by its own norm and reaches optimal regret without knowing the payoff scale. On zero-sum matrix games it converges linearly in the last iterate, and the authors attribute this to norm saturation, where the regret norm rises to a finite limit and freezes the step size. They prove this always happens and that the last-iterate Nash gap vanishes on every matrix game. Near a unique strictly complementary equilibrium they give a closed-form linear rate and an observable ratio certificate. ★★★★ Notable: Explains why scale-invariant regret matching converges linearly in zero-sum matrix games via norm saturation, giving explicit bounds, a closed-form rate, and a ratio certificate for monitoring.

An information identity showing the geometry behind rare deviation events

2026-09-30T00:00:00+09:00

An information identity reveals the geometry of deviation events arXiv · Sep 30, 2026The paper studies the probability of a deviation event, where the empirical measure of n independent draws falls in a set of distributions. At every n, the exponent of this probability splits into two nonnegative terms: a rate, equal to n times the relative entropy of a typical draw's distribution under the event relative to the population, and a dependence term among the draws, measured by total correlation. The relative sizes of the two terms depend on the geometry of the set, which the paper examines. ★★★★ Notable: Gives an exact split of deviation-event exponents into a relative-entropy rate and a total-correlation dependence term, showing how their balance reveals set geometry, with concrete results for single

A transmitter design that adapts to the receiver for a 200 Gbps C-band optical access network.

2026-09-30T00:00:00+09:00

Receiver-Informed Transmitter for 200Gbps C-Band Direct-Detection PON With FFE-Free ONU arXiv · Sep 30, 2026This paper proposes and numerically demonstrates a receiver-informed transmitter (RITX) waveform design for 200 Gb/s PAM-4 C-band direct-detection passive optical networks. Using 50 GHz optoelectronics, RITX reaches N1/E1 loss classes with an FFE-free ONU and supports a 40 km reach. At 1:64 and 1:128 splits, it yields network-wide complexity savings of 48.1% and 66.7%, respectively.

Simple Mechanisms in Matching Markets: A Tight e-Approximation for Gains from Trade

2026-09-30T00:00:00+09:00

The Power of Simple Mechanisms: A Tight e-Approximation for Gains from Trade in Matching Markets arXiv · Sep 30, 2026This paper studies how well simple mechanisms approximate gains from trade (GFT) in two-sided matching markets with independent buyer values and seller costs. It shows that the Generalized Random-Offerer mechanism achieves an exact worst-case approximation ratio of e relative to first-best GFT, improving the previous 3.15 bound. The authors also construct markets with regular distributions and a single knapsack constraint showing this guarantee is tight. ★★★★★ Notable: The abstract reports that GRO's exact worst-case approximation ratio to first-best gains from trade is e, improving the earlier 3.15 guarantee, with the authors showing tightness via regular-distribut

How GPU server layout, parallelism and congestion control jointly affect MoE inference speed

2026-09-30T00:00:00+09:00

Joint Effects of GPU Server Topology, Parallelism, and Congestion Control on MoE Inference: A Controlled Simulation Study arXiv · Sep 30, 2026This paper uses ASTRA-sim with an NS-3 backend to simulate MoE model inference across 32 GPU ranks under controlled settings. It tests six server topologies, four TP/EP partitions, two TP collective algorithms and four network congestion-control modes, giving 768 deterministic simulations. The authors find that exposed communication accounts for 89.9% to 95.8% of mean completion time, and that topology, parallelism, collective implementation and congestion control jointly determine completion time.

Innovation-Based Sampling: Conveying Information Through the Direction and Timing of New Information

2026-09-30T00:00:00+09:00

Innovation-Based Sampling for New Information arXiv · Sep 30, 2026This paper introduces innovation-based sampling of continuous stochastic processes, generating a sample only when the process reveals sufficiently new information relative to earlier samples. For a Wiener process, it characterizes the direction, timing, and frequency of innovations, showing they become sparser over time so the sampling rate vanishes asymptotically. The paper also studies remote estimation from sparse innovations, finding that the absence of an innovation carries information and yields substantially lower mean-square error than a conventional zero-order hold estimator.

Emulating FP64 matrix products with low-precision INT8, FP8 and FP4 operations

2026-09-29T00:00:00+09:00

FP64 Is All You Want, INT8 Is All You Need, FP4/6/8 Is All You Have arXiv · Sep 29, 2026The paper treats Ozaki scheme II and its FP8 and FP4 variants as one family that emulates FP64 matrix products using residues modulo coprime moduli. It frames choosing a scheme as a combinatorial problem that minimizes the number of low-precision GEMMs and derives lower bounds on that count. Implemented on three Blackwell GPUs, the INT8, FP8 and FP4 schemes run faster than native FP64, up to 83x on B300. ★★★★★ Notable: Unifies Ozaki-style schemes as a combinatorial optimization with GEMM lower bounds, yielding first FP6 schemes; the authors report INT8, FP8 and FP4 schemes outperform native FP64 on three Blackwell G

Optimal codes that correct up to t deletions in permutations

2026-09-29T00:00:00+09:00

Order-Optimal Systematic Permutation Codes for Correcting t Deletions arXiv · Sep 29, 2026This paper constructs systematic permutation codes that correct multiple deletions under two models: symbol-invariant deletions (SIDs) and permutation-invariant deletions (PIDs). For any fixed t≥1 and large n, the encoders insert redundancy symbols while preserving the order of the message symbols. The redundancy is (7t−1)log n+O_t(1) bits for PIDs and 4t log n+O_t(1) bits for SIDs. Encoding and decoding run in n^{O(t)} time.

MEDEM: A Design Method for Multi-Engine Deep Learning Accelerators

2026-09-29T00:00:00+09:00

MEDEM: Multi-Engine DL Accelerator Design Methodology arXiv · Sep 29, 2026This paper proposes MEDEM, a systematic methodology for designing deep learning accelerators that combine multiple complementary engines. It co-designs candidate engines and selects the combination that minimizes aggregate execution cost under a resource budget. The authors report geometric-mean improvements of up to 4.84x in energy-delay product and 1.59x in throughput over state-of-the-art designs.

[Correction] Accelerated Dataflow: Scaling Efficient Inference in Streaming Data Analytics

2026-10-30T17:04:41+09:00

This item has left this feed: Moved from "Other Computer Science" to "AI & Machine Learning" Accelerated Dataflow: scaling inference in streaming data analytics with efficiency arXiv · Oct 16, 2026