Crypto & Blockchain · Mon, Nov 9, 2026

Bitcoin, Ethereum, stablecoins, DeFi, smart contracts, scaling and consensus.

RSS feed
arXiv cs.MA · cs.CL

Agent4Crypto: A Dual-Reflective and Market-Feedback-Aligned LLM-Based Multi-Agent System for Cryptocurrency Trading

In briefAgent4Crypto: An LLM-Based Multi-Agent System for Cryptocurrency Trading with Dual Reflection and Market Feedback

SummaryThis paper presents Agent4Crypto, a multi-agent system for cryptocurrency trading in which specialized agents handle data collection, analysis, risk assessment, reflection, and trading. Three trader agents with aggressive, neutral, and conservative styles propose actions, and a PPO-based gating policy learns how much weight each trader receives, adjusted through two reflection processes. Experiments on BTC, ETH, and XRP report improved total return and Sharpe ratio compared with traditional, machine learning, reinforcement learning, and existing LLM-based trading baselines.

Guangyan Gan, Jiahao Li, Jiashun Wang et al.

arXiv cs.DC · cs.LG

DeFSM: Robust and Efficient Decentralization of Federated Submodular Maximization

In briefDeFSM: A decentralized, Byzantine-robust way to run federated submodular maximization

SummaryThe paper addresses federated submodular maximization, where a central server is a single point of failure and some clients may be Byzantine (malicious). It proposes DeFSM, which replaces the server with a lightweight smart contract and a pool of consensus-free independent agents. Its aggregation rule, SafeGM, uses Tukey depth and safe points from combinatorial geometry to make robust decisions on 2D Gaussian random projections, then lifts them back to convex combinations of actual updates. On CIFAR10 and PathMNIST with up to 30% Byzantine clients and agents, DeFSM never collapses and retains about 91% of the ideal reference quality on average.

Duc A. Tran, Dung Truong, Duy Le

arXiv cs.DC

A Decentralized Quantum-Resistant Notary Scheme for Blockchain Interoperability with a Novel Dynamic Trust Model

In briefA quantum-resistant, decentralized notary scheme for blockchain interoperability with a dynamic trust model

SummaryThe paper addresses two weaknesses of existing blockchain notary schemes: vulnerability to quantum attacks and rigid trust management. It proposes a decentralized framework combining ML-KEM, ML-DSA and AES-256, using asymmetric cryptography for authentication and symmetric encryption for bulk data. Under parallel loads of up to 270 transactions per second, the end-to-end performance gap with elliptic-curve approaches averages under 6%. A dynamic trust model derives an adaptive discount factor from the variance of a Verifier's behavior using Popoviciu's inequality, and was evaluated over 200 epochs under adversarial conditions.

Koosha Esmaeilzadeh Khorasani, Saeed Moradi, Sara Rouhani

arXiv cs.DC

Verifiable, Privacy-Preserving and Collaborative Multi-Objective Optimization with NSGA-II on Blockchain

In briefVerifiable, privacy-preserving collaborative multi-objective optimization on blockchain (zk-NSGA-II)

SummaryThis paper proposes zk-NSGA-II, a protocol that runs NSGA-II on the RISC Zero zkVM so that each evolutionary generation is certified by a succinct proof verified on-chain. Solvers keep private off-chain populations and contribute non-dominated solutions to a shared on-chain Pareto front, with hash chains linking generations to prevent fabricated contributions. The protocol is evaluated on twelve ZDT and DTLZ benchmark problems and a Hang Seng dataset portfolio optimization case study across nine metrics. The results indicate that trustless collaborative multi-objective optimization is feasible with current zero-knowledge and blockchain technologies.

Goshgar Can Ismayilov