Tight limits on fair AMM mechanism design under weak efficiency
WhyProves impossibility of IC and wLE for deterministic anonymous batch AMMs under plain consensus, with tight characterization of strategy restrictions.
Who Patches the Lightning Network? Release-line analysis of Lightning gossip, 2023–2026
WhyInfers release lines from gossip feature bits, validated on daily snapshots, showing platform packaging drives Lightning patching more than disclosures.
Sybil-proof adaptive clinching auctions and their use in blockchain market design
WhyProves Sybil-robust incentive compatibility for adaptive clinching and designs a two-sided AMM, backed by impossibility results and proofs with broad DeFi relevance.
PQ-Vault: A provably quantum-safe way to hold Ethereum assets without changing Ethereum
WhyProves unmodified Ethereum cannot protect vulnerable accounts without quantum-safe vault release, and finds compact certificates drop to about 60 bits.
Tritone: Pairwise Garbled Disputes for Bitcoin Groth16 Proof Checks
WhyGeneralizes two-party BABE to N operators with a new Lamport-adaptor binding gadget, verifying Groth16 on Bitcoin without consensus changes, with implementation and on-chain cost measurements.
Isotaxis: Optimal Asynchronous Byzantine Consensus with Ordering Linearizability
WhyFirst asynchronous SMR achieving ordering linearizability with optimal communication and Byzantine resilience, validated by experiments on 4 to 100 nodes.
Stars are a rough AI assessment based on the title and abstract: each paper is scored for novelty, strength of evidence and breadth of impact, then mapped to 1–5 stars (about 2% get five stars and 10% get four). The site only marks four- and five-star papers as Notable, with a one-line reason; no mark does not mean a paper is weak. The full text is not read, so treat the stars as a reading aid, not a verdict on quality. "Code" means the abstract states that code, data or tools are released.
In briefUsing quantum annealing to detect and locate gamma-ray sources in AGILE data
SummaryThis paper reformulates maximum-likelihood gamma-ray source detection and localization as a Quadratic Unconstrained Binary Optimization (QUBO) problem and solves it with simulated quantum annealing. The method is applied to simulated observations from AGILE's Gamma-Ray Imaging Detector in the 30 MeV to 50 GeV band, generated with AGILEpy. The flux and position estimates agree with classical results within statistical uncertainty, and the correct grid node is identified in 97% of the maps.
Alessandro Rizzo, Nicol\`o Parmiggiani, Andrea Bulgarelli et al.
In briefUniPCH: A Unified Framework for Cross-Chain Payments via Payment Channel Hubs
SummaryThis paper presents UniPCH, a two-layer framework for private cross-chain payments across heterogeneous blockchains. The lower UniPC layer defines common channel state transitions and security requirements, with realizations for UTXO and smart-contract account ledgers. The upper layer introduces selective re-signing blind conditional signatures (S-RBCS) to coordinate conditional authorization across hubs, enabling a four-phase payment protocol. The authors report security properties and a Go prototype evaluation of running time and bandwidth.
In briefUndisclosed Fees in Hook-Based Decentralized Exchanges
SummaryThis paper presents a large-scale empirical study of Hook contracts on Uniswap v4 and PancakeSwap Infinity across Ethereum and BNB Smart Chain. The authors analyze 7,873 Hook deployments and 50,011 pools using bytecode analysis to identify settlement charges not itemized in standard fees. They find that 24.30% of Hooks can impose such deductions, which occur in 59.23% of relevant swaps and generate substantial revenue. The results suggest a gap between displayed fees and actual trading costs, motivating explicit disclosure.
In briefTight limits on fair AMM mechanism design under weak efficiency
SummaryThis paper studies miner extractable value (MEV) in automated market makers (AMMs), where block builders can profit by controlling transaction order. It proves that no deterministic, anonymous batch AMM can be both incentive compatible and weakly locally efficient under plain consensus without fair sequencing, even with censorship resilience and unlimited block space. The impossibility relies on Sybil identities, over-reporting budgets, and misreporting values. Removing any one of these strategic capabilities makes both properties achievable under censorship resilience.
WhyProves impossibility of IC and wLE for deterministic anonymous batch AMMs under plain consensus, with tight characterization of strategy restrictions.
In briefSybil-proof adaptive clinching auctions and their use in blockchain market design
SummaryThis paper examines whether the adaptive clinching auction for budget-constrained bidders stays robust when a bidder splits its budget across fake identities. The authors propose a generalized adaptive clinching framework and prove that bidders gain nothing by splitting into any number of identities, as long as their total reported budget does not exceed their enforceable true budget. They extend the mechanism to a two-sided automated market maker for decentralized finance and show, through impossibility results, that these guarantees are essentially the best possible.
WhyProves Sybil-robust incentive compatibility for adaptive clinching and designs a two-sided AMM, backed by impossibility results and proofs with broad DeFi relevance.
In briefFIDES-Live: Checking Whether Trading AI Agents Behave Consistently
SummaryThis paper presents FIDES-Live, an extension of the FIDES framework for auditing sequences of trading agent actions, orders, execution receipts, and reports. It assesses decision validity, adherence to numerical commitments, reconciliation with execution and ledger records, and report accuracy grounded in receipts. The authors retrospectively audited ten inference routes over 7,299 cryptocurrency account-rounds and a 310-round equity pilot. The paper states that its findings do not establish long-run trading skill, faithfulness of explanations, or causal benefits of audit feedback.
In briefPhysical structure as an organizing principle for universal Hamiltonian learning
SummaryThis paper introduces UniTB, a universal E(3) equivariant framework that combines density-functional tight-binding physics with deep learning to predict electronic Hamiltonians across many chemistries. Trained on 29,242 crystals, it reaches a 0.48 meV matrix-element error on nonmagnetic test crystals spanning 68 elements. A faster variant, UniTB-Flash, is also presented, and both models are used to screen 16,195 crystals.
Mingkang Liu, Zhanghao Zhouyin, Guanghui Wang et al.
In briefFuji: How Demand Bursts Take Shape, What They Consist Of, and How They Spread on Robinhood Chain
SummaryThis paper measures sudden surges in demand on a blockchain marketplace, using Robinhood Chain, a layer-2 network, over its first three active months and more than 50 million Uniswap swaps. During bursts, traders touch about as many pools as in quiet periods, mostly different ones, and trade more intensively, with median peaks 2.5 times normal load. The rise and fall of demand are described by two exponential functions, and the authors compare two explanations for bursts, finding externally triggered cascades fit the data better than self-excitation.
In briefWeekend Crypto Volatility and Monday Risk: Whether weekend crypto price swings predict Monday variance in traditional markets.
SummaryThis paper studies whether volatility in cryptocurrencies such as Bitcoin, Ether and XRP, which trade through the weekend, predicts Monday variance in the S&P 500, gold and a US Treasury ETF. The authors add a common weekend crypto factor to heterogeneous autoregressive models using daily data from 2017 to 2026. Predictive content is significant after 2024 and reduces out-of-sample losses, though statistical evidence varies across markets.
In briefMerkle-tree-based public auditing of cloud object storage using Ethereum smart contracts
SummaryThe paper proposes a public auditing system for cloud object storage in which the hash and name of each object form leaves of a Merkle tree. Only the tree's root is stored in an Ethereum smart contract, while the objects remain with the provider. An auditor downloads an object and checks it against the stored root using a short Merkle proof to mark it as intact, modified or missing. In experiments, the system detected all single and multiple modifications and deletions with no false alarms, and checking one object took about 10 ms regardless of dataset size.
Siri Reddy Tadi, K. Gangadhara Rao, M. Venu Gopalachari et al.
In briefIsotaxis: Optimal Asynchronous Byzantine Consensus with Ordering Linearizability
SummaryThis paper presents Isotaxis, a state machine replication protocol with ordering linearizability, designed to resist adversaries who manipulate transaction order, such as in MEV attacks. Its communication complexity is O(nℓ+λn²), which is optimal when ℓ≥λn, and it is described as the first asynchronous protocol to achieve ordering linearizability. Experiments on up to 100 EC2 instances show it outperforms Pompē at all tested scales from 4 to 100 nodes.
WhyFirst asynchronous SMR achieving ordering linearizability with optimal communication and Byzantine resilience, validated by experiments on 4 to 100 nodes.
In briefBeyond Wasserstein Proximity: Finding Better Transfer Sources
SummaryThis paper examines how to choose pretraining data sources for transfer, arguing that the most similar data is not always the most useful. The authors test five-day volatility forecasting for ten Binance spot cryptocurrencies across six model families. Selecting farther Gaussian-Wasserstein sources improved persistence-relative skill for several models. The findings are presented as exploratory results on a fixed historical dataset.
Gordei Verbii, Georgii Killinkarov, Vladimir Yafarov
In briefPerformance guarantees for a permissionless lending protocol
SummaryThis paper proposes a lending protocol that the authors say carries far fewer risks than pool-based protocols such as Aave and Compound. Lenders and borrowers make competing interest-rate offers, and loans carry a fixed rate for their full term. The authors model collateral value as a random walk and use characteristic functions and Fourier inversion to derive exact expressions for liquidation and repayment-shortfall probabilities, illustrated with Bitcoin collateral.
In briefDecentralized medical data sharing system using off-chain storage and view-based access control
SummaryThis paper proposes VASCO, a decentralized medical data sharing system that does not rely on a central server, a specific cloud storage such as IPFS, or a secure channel. It encrypts both data and metadata end-to-end, using eight chaincodes built on view-based access control. The authors report scalability tests on a 21-peer blockchain and interoperability checks across 30 combinations of standards, formats, and protocols.
In briefBE-CTDE: Blockchain-based multi-agent learning with anomaly detection and token incentives
SummaryThis paper proposes BE-CTDE, a framework that uses blockchain as a training-management layer for multi-agent reinforcement learning. A behavioral verifier compares executed trajectories with shared-policy recommendations, and token mechanisms (slashing, burning, rewards, compensation) adjust each agent's influence on the global policy update. In multi-agent path finding experiments, it achieves higher success rates under attack, even when adversaries are a majority, with collision rates up to 35% lower than detection without economic consequences and up to 43% lower than conventional stake slashing.
Chih-Wei Huang, Hsin-Chieh Lee, Yun-Chen Yu et al.
In briefPost-quantum IoT blockchain for securing edge and physical AI agents in critical systems
SummaryThis paper addresses the risk that quantum computers could break public-key authentication used in distributed cyber-physical systems and edge AI. It introduces rotational secrecy, an architectural goal that limits key exposure through authenticated transitions between cryptographic epochs, and presents PARS, a mechanism for coordinating key establishment, rotation, and reconciliation among agents. It also describes PARS-ESL, an optional permissioned ledger that records only security state such as key epochs, membership, and revocation, not high-volume operational data. The authors conclude that ledger-assisted post-quantum key-epoch management is feasible for selected edge AI and CPS security-control workloads under stated assumptions.
In briefMore Meters or More Prices? Comparing Single and Multi-Dimensional Ethereum Fee Designs
SummaryThis paper examines Ethereum's shift from a one-dimensional transaction fee mechanism to a multi-dimensional one that prices execution, data, and state creation separately. The authors build a demand model calibrated from 2025 block-capacity increases and find that state creation is the most price-responsive resource. Combining this model with transaction-level metering reconstruction and block-level simulation, they report that the multi-dimensional design removes fee coupling and yields more execution throughput at similarly low congestion.
Fei Wu, Maria In\^es Silva, Stefanos Leonardos et al.
In briefGraphene microwave calorimeter with zeptojoule sensitivity and nanosecond recovery
SummaryThe study calibrates a graphene Josephson-junction microwave calorimeter using nanosecond pulses and independently measures the thermal parameters of the same device. At 110 mK, the calibrated pulses give a switching-energy scale of 18.1 meV (2.90 zJ), and two-pulse thermometry gives an intrinsic relaxation time of 19.5 ns. The measured electronic heat capacity and electron-phonon coupling are larger than ideal estimates, and a thermal model using these measured values predicts 21.2 meV, matching the observed switching response.
In briefQuantum Cadence: a post-quantum consensus protocol for multiple concurrent proposers that avoids certificates
SummaryMost high-performance blockchains rely on quorum certificates built from aggregated signatures, which large-scale quantum computers could break, and no post-quantum scheme matches their size and aggregation speed. This paper presents Quantum Cadence, in which each validator counts quorums from the messages it has received, so no certificates are needed. Its fast path keeps the optimal latency of Chorus, and when that fails it falls back to Coda, a signature-free Byzantine agreement. In simulations with 200 validators, its latency is comparable to Chorus and it scales better as the number of proposers grows.
Mohammad Mussadiq Jalalzai, Tobias Klenze, Jovan Komatovic
In briefTritone: Pairwise Garbled Disputes for Bitcoin Groth16 Proof Checks
SummaryThis paper presents Tritone, a protocol that lets Bitcoin enforce the outcome of verifying a Groth16 proof without changing consensus rules. It extends the two-party BABE protocol to a fixed committee of N operators, and the locked funds stay secure while at least one operator is honest. The authors formally prove security against up to N-1 corrupted operators and implement the protocol in Rust. They report transaction sizes on Bitcoin mainnet and local storage measurements.
WhyGeneralizes two-party BABE to N operators with a new Lamport-adaptor binding gadget, verifying Groth16 on Bitcoin without consensus changes, with implementation and on-chain cost measurements.
In briefBeyond Copying: Lifelong Smart Contract Security Task Construction via Protocol Mutation
SummaryThis paper proposes CONSTRAINTDIAL, a framework that automatically builds smart contract security tasks by mutating protocols, to test AI agents on detecting, exploiting, and patching vulnerabilities. It changes the calls or inputs needed for an attack so the original fails while exploitability and normal functionality are preserved, and new incidents are added over time. Using 203 public cases, the authors built 3,045 task inputs. Protocol mutation lowered success with supplied public answers by 10.9 percentage points on average compared with reconstruction alone.
In briefTracing Money Laundering Across DeFi, Bridges and Mixers on Ethereum
SummaryThis paper presents FPTracer, a framework for tracing illicit assets through Ethereum services such as DeFi protocols, cross-chain bridges and mixers. It recovers intra-transaction asset-flow paths using a Context-Aware Transfer Tree, classifies transactions with cross-chain intent, and analyzes post-mixer activity using upstream laundering evidence. Evaluated on 20 real-world cases, it achieves the highest recall for known laundering addresses and destinations and finds 30 previously unreported bridge uses.
In briefA systematic survey of optimization tools for Hydra multi-party channels
SummaryThis paper is a systematization of knowledge (SoK) survey on blockchain Layer-2 scaling. It reviews existing two-party payment channel protocols and the tools built on them, explaining their core ideas. It then identifies use cases that call for new protocols for Hydra's multi-party channels, lists open problems, and suggests existing tools that could address them or serve as starting points.
In briefSILMARILS-MV: An Information-Theoretically Secure Multi-Verifier Signature Scheme for Permissioned Blockchains
SummaryThe paper proposes SILMARILS-MV, a multi-verifier signature scheme that remains secure against unbounded adversaries and needs no trusted dealer for key issuance. Signatures can be linearly aggregated, and the committee can sign threshold certificates of constant size. According to the authors' analysis and HotStuff prototype, it reduces authentication traffic and cryptographic CPU cost compared with ML-DSA-44.
Khadijeh Bagheri, Hassan Khodaiemehr, and Chen Feng
In briefDelta N-gram: Forward-Only Writing of New Information into a Language Model's N-gram Memory
SummaryThis paper proposes Delta N-gram, a method that writes new information into a language model's native hashed N-gram table using forward passes only, leaving the pretrained table unchanged. Updates go into a sparse patch via a small writer that outputs targets and gates, combined with a block-averaged delta rule. Experiments report document-specific memory, better stability over many writes than the plain writer, and drift in some frequently read format rows.
In briefDLT Terms 100: A Blockchain Glossary Based on ISO 22739:2024
SummaryThis paper proposes "DLT Terms 100," a glossary intended to improve understanding of blockchain, or distributed ledger technology (DLT). The authors note that gaps between business and technical terminology, and confusion from existing financial terms being reused, hinder comprehension. The glossary is based on ISO 22739:2024 and categorizes terms by how often they appear in representative domestic and international literature and by difficulty level.
In briefTemporal liquidity matching for blockchains: an EIP-1559 extension admitting transactions below the base fee
SummaryThe paper proposes temporal liquidity markets (TLM), in which transactions seeking early block execution (consumers) fund transactions that accept later execution (providers). In a block-local EIP-1559 extension, providers can be admitted below the base-fee floor, and their gas is excluded from the ordinary demand controller input. The authors state an adversary model with strategic senders and builders and prove eight properties, including validator-checked rule validity, full burning of the base fee, and no subsidy.
In briefCambium: Shared Commitment Roots Built by Participants
SummaryThis paper presents Cambium, a permissioned peer-to-peer service in which participants jointly build each cycle's commitment root, so no single party decides which inputs it includes. Fixed-size committees certify inputs, and a recursive-doubling merge spreads root construction across all nodes. A Go prototype on AWS completed cycles in 12 to 148 ms, verified proofs in under 1.5 ms, and showed logarithmic growth.
Mansoor "Manny" Ahmed, Dann Toliver, Ekaterina Pavlova
In briefWho Patches the Lightning Network? Release-line analysis of Lightning gossip, 2023–2026
SummaryThe paper infers which software release line each Lightning node runs by matching the feature bits it announces in public gossip against implementation sources. Analyzing 814 daily graph snapshots from 2023 to 2026, it finds nodes upgrade on release and platform schedules rather than after security disclosures. The authors suggest that platforms packaging releases are the main lever for nodes that lag behind.
WhyInfers release lines from gossip feature bits, validated on daily snapshots, showing platform packaging drives Lightning patching more than disclosures.
In briefUsing time-series foundation models to detect financial crime
SummaryThis paper proposes TCEE, a framework that summarizes each account's transaction history with a frozen time-series foundation model and adds the relevant temporal context to each transaction's representation. These representations can be used for edge classification or as edge features in a graph neural network. Across six AMLWorld datasets, TCEE improves PR-AUC by 10.8% on average over the strongest baseline on five datasets and trains 6.0 times faster end-to-end.
Danae Natalie Savvidi, H. \c{C}a\u{g}r{\i} Bilgi, Kubilay Atasu et al.
In briefSecure metaverse computation offloading with smart update admission and adaptive Byzantine redundancy
SummaryThis paper studies secure computation offloading for latency-sensitive metaverse services in computing-in-the-network (COIN) systems. It jointly decides which shared-state updates to commit to a blockchain, how much PBFT Byzantine redundancy to provision, and where to run tasks. The authors propose a hierarchical R-PBFT approach combining an identical-interest game with a feasibility-filtered double deep Q-network. Numerical results report reward gains over the strongest baseline and ablations that assess feasibility filtering and adaptive deferral.
In briefHow Not to Be Censored: Guaranteeing Client Transaction Inclusion in Blockchain Consensus
SummaryThis paper observes that blockchain consensus is typically client-server, with clients only submitting transactions and relying on validators to include them. The authors propose a leaderless consensus protocol in which clients actively participate and define inclusion safety, a guarantee that a client's transaction enters some block produced in an epoch. The analysis covers synchrony and asynchrony, showing that client or replica connectivity can substitute for one another, though relying on replica connectivity requires a longer epoch.
In briefPQ-Vault: A provably quantum-safe way to hold Ethereum assets without changing Ethereum
SummaryThe paper warns that a quantum computer could recover the private key of any Ethereum account whose public key is already on chain. The authors prove that sidechains or bridges cannot protect such accounts without changing Ethereum's rules, so funds must move into a vault contract released only through quantum-safe primitives. They present PQ-Vault, combining this vault, a quantum-safe proof-of-stake sidechain, and an optimistic dispute protocol.
WhyProves unmodified Ethereum cannot protect vulnerable accounts without quantum-safe vault release, and finds compact certificates drop to about 60 bits.
In briefRegime-aware multi-asset portfolio allocation using hidden Markov models and nested clustered optimization, tested on Indonesian assets.
SummaryThe paper proposes a tactical asset allocation framework that combines Gaussian Hidden Markov Model regime detection with Nested Clustered Optimization, with regimes identified out of sample. It is tested on an Indonesian multi-asset universe including government bonds, LQ45 equities, gold, CNY/IDR exchange rates, and Bitcoin over 56 out-of-sample months from 2021 to 2025. The regime-aware version limits maximum drawdown to 4.44%, but the regime-agnostic NCO shows a lower drawdown of 2.19% and higher cumulative return. The authors find that a rolling 24-month HMM switches regimes in 55 of 56 months, causing regime drift that erodes the expected gains.
In briefWhy Ethereum-style SDK defaults fail on UTXO-style chains: three reusable design patterns
SummaryThis paper argues that mainstream Web3 SDKs such as ethers.js and viem assume an account-based ledger, a single secp256k1 signer, and transactions built without input selection. These assumptions break on UTXO-family chains like Nervos CKB's Cell model and when wallets use different cryptographic primitives. Using the TypeScript SDK CCC, the authors extract three recurring patterns from its release history and exported APIs: a declarative transaction composer, a unified signer abstraction, and a drop-in connector UI component. The paper presents these as transferable design lessons for SDKs serving UTXO-family ledgers and heterogeneous wallets.
In briefMise-en-Video: Fine-Grained 3D Control for Multi-Shot Video Generation
SummaryThis paper presents Mise-en-Video (MeV), a framework that lets creators stage multiple shots in 3D space and then generate the matching multi-shot video. It combines a hierarchical attention mechanism for narrative consistency across levels with 3D-aware layout guidance for precise per-shot spatial control. The authors also built a 160K-video dataset with hierarchical multi-shot texts and oriented 3D bounding boxes for training and evaluation. Experiments report that MeV outperforms strong baselines in video quality, inter-shot consistency, and subject-level controllability.
In briefA blockchain consensus method where miners run AI inference instead of hash puzzles
SummaryThis paper proposes Proof-of-ML-Inference (PoML), a consensus protocol in which miners perform user-requested machine-learning inference instead of computing hash puzzles. The authors formalize the computational assumptions needed for ML inference to replace hash work and argue that LLMs and diffusion models satisfy them. They analyse security by reducing it to the Bitcoin Backbone model and evaluate block-time stability in a simulated environment.
In briefFortress: A Fork-Resistant Transparency Log Using Social Graph Structure
SummaryTransparency systems such as Key Transparency rely on external anchors like bulletin boards, blockchains, or out-of-band gossip among auditors to ensure everyone sees the same directory. This paper proposes using the social graph formed by users themselves as a replacement, formalizing a primitive called sg-FORT and constructing the Fortress protocol. In honest runs, out-of-band interactions occur only when users reset their secrets, and security is proven under standard cryptographic assumptions. The authors present this as a feasibility result under simplifying assumptions such as synchronous epochs and always-online users.
In briefTesting Whether AI Trading Agents Follow Their Rules or Just Get Lucky
SummaryThis paper tests whether four AI trading agents followed their declared strategies and risk rules. The authors analysed 2,627 closed positions, mostly simulated against live prices with a 0.09% round-trip fee, and 18% using real Hyperliquid orders. The fleet finished net positive after fees with high rule adherence, and the study provides a toolkit to separate rule-following from luck. The authors stress these are trade-level results, not proof of account returns or a lasting edge.
In briefPeer-relative trust screening to resist label-flipping attacks in semi-decentralized federated learning
SummaryThis paper proposes a peer-relative trust-based update screening method to make semi-decentralized federated learning more robust against label-flipping attacks. It combines a robust validation-utility score with direction- and norm-based update quality measures. In experiments with 50 MNIST clients and 20% persistent attackers, the method reached 98.37% accuracy under moderate heterogeneity, rejecting 95.82% of malicious updates, and 95.25% accuracy under severe heterogeneity, rejecting 88.55% of malicious updates.
In briefOn Pearl Blockchain's Proof of Useful Work: Systematization and Flexibility
SummaryThis paper revisits the proof-of-useful-work (PoUW) scheme from the Pearl blockchain white paper. It offers a systematic sensitivity characterization as an alternative to the Jackpot policy, along with an efficient algorithm for verifying the sensitivity condition. Building on this, it proposes a more flexible PoUW scheme in which each miner independently chooses its noise magnitude to balance computational accuracy and mining efficiency.
In briefBlockchain participation games with upfront contribution commitment
SummaryThis paper studies blockchain participation games where entry is costly, contributors are randomly selected, and success requires reaching a threshold. It characterizes all pure Nash equilibria under homogeneous selection and analyzes how heterogeneous selection affects who commits. The authors show a minimum-outlay reward scheme that supports full contribution as a Nash equilibrium, and note that requiring zero registration reward can raise the minimum outlay under heterogeneous selection.
In briefEmpirical Study of TEE-Based Rollup Sequencer Designs
SummaryThis paper experimentally evaluates designs that use trusted execution environments (TEEs) to strengthen Ethereum Layer 2 rollup sequencers against censorship and manipulation. It compares separating proposer and builder logic (PBS) into independent services with a monolithic sequencer design. The experiments find that TEEs have a small impact on rollup performance, and PBS does not significantly amplify that impact. The authors also report that PBS can yield significant performance gains through more specialized, efficient builder implementations.
Claudiu-Nicu B\u{a}rbieru, Corentin Libert, Peterson Yuhala et al.
In briefA decision engine for choosing where to deploy smart contracts
SummaryThis paper proposes a systematic method to help choose among Ethereum layer 2 (L2) blockchains for deploying smart contracts. It uses a set of metrics and user-defined profiles that reflect different priorities among them. The authors collect major smart contract categories from a community-maintained source and six months of transactions from Arbitrum, Optimism, and Base, then compute metrics and scores per category and chain.
Oana-Otilia Captarencu, Emanuel Onica, Andrei Arusoaie
In briefABF-Bench: A Verifiable Benchmark for Agentic Blockchain Forensics
SummaryThis paper introduces ABF-Bench, an interactive benchmark that tests whether AI agents can trace cryptocurrency funds using public ledger records, requiring each claimed transfer to cite records the agent actually retrieved. The authors also propose the ABF agent harness, built around a working memory, and compare it with general-purpose agents and fixed-order tracers. The harness reaches 75% edge recall, versus 47% for Codex and 10% for Claude Code. The paper also notes that agents often stop early and lose recall on larger tasks.
In briefClassifying DeFi transaction methods using network graphlet patterns
SummaryThis paper examines whether network topology can identify the transaction methods that describe Decentralized Finance (DeFi) operations. The authors extract directed graphlet features to classify the 30 most frequent methods. They report that graphlet features complement event log counts and improve classification performance. Using SHAP (SHapley Additive exPlanations), they derive model explanations that reveal method-specific graphlet structural patterns.
In briefPaper on whether adaptive transaction fees stay stable across multiple blockchains
SummaryThis paper studies whether adaptive base fee mechanisms like Ethereum's EIP-1559 remain stable when several blockchains are linked by cross-chain transfers. The authors model multi-chain fee markets as a nonlinear dynamical system with delayed feedback and identify delay-amplified over-correction as the root cause of instability. Experiments show that cross-chain coupling shrinks the safe range of update rates and that instability spreads to peer chains. The authors also propose a lightweight protocol-level mitigation that removes the over-correction.
In briefMarket Efficiency in a Changing Asset Population: Mobility, Entry, and Survival Across Cryptocurrency History
SummaryThis paper tests symmetry restrictions implied by weak-form efficiency on the Markov chain of cryptocurrency capitalisation movements, using 8,970 listed assets across 690 weekly cross-sections from 2013 to 2026. Symmetry is rejected in every era, but asymmetry falls, and the convergence is attributed to entry-driven compositional churn rather than changes in coins' own valuations. The bottom decile shows 27.5% weekly returns on paper but none among tradable assets, with capacity of $293,000.
In briefQuarkShard: Splitting Cross-Shard Requests into Single-Shard Transactions
SummaryThis paper proposes QuarkShard, a sharding protocol for blockchains that splits cross-shard user requests into transactions that each need consensus within only one shard. A request decomposition algorithm is said to preserve fund or location atomicity, and the design is said to improve resistance to MEV and flooding attacks. In tests across four AWS regions, the authors report a peak throughput of 1024k requests per second with latency under 4 seconds, and up to 210.7% and 314.8% of the maximum throughput of Kronos and 2PC.
In briefProvable Authority: A Verifiable Protocol for Checking Organizational Authority Before Transactions
SummaryThe paper proposes the ChainIT Authority Protocol, which lets a counterparty or payment rail independently verify, before execution, that the party presenting a transaction holds current organizational authority to execute it. It separates identity, a versioned authority record, sourced evidence, policy predicates, and execution credentials, binding decisions and human approvals to a canonical transaction digest. The paper also covers an extension for AI agent subjects, enforcement profiles for fiat and stablecoin rails, and stated residual risks and non-goals.
