smart-contracts

This paper proposes a frequency-aware continual learning framework using large language models for smart contract vulnerability detection, addressing adaptation, forgetting, and consolidation challenges with efficient methods like FA-LoRA and APPM.

ArcBounty is an open-source marketplace where AI agents can autonomously take paid jobs, complete them, and get paid via smart contracts without human intervention, with on-chain reputation vouching for agents.

The paper proposes a neurosymbolic framework that lifts LLM outputs into typed compositional derivations using Combinatory Categorial Grammar, enabling structural checking and early hallucination detection.

AI-powered security tools like Mythos are making smart contract audits cheaper and faster, potentially shifting industry standards for security due diligence. While AI can quickly find coding flaws, experts note it cannot replace human judgment or prevent losses from social engineering and operational failures.

This paper introduces SCOPE-FL, a hierarchical federated learning framework that uses the Top Trading Cycle algorithm to ensure strategy-proofness and Pareto efficiency in client selection, with reward distribution via Shapley value approximation and blockchain-based execution.

This paper identifies a new class of attack called Ghost Fills in Polymarket's hybrid off-chain/on-chain architecture, where attackers exploit the time gap between matching and settlement to revert orders, leading to over $1.49M in illicit profits and putting $1.78B at risk.

AI agents can now interpret ambiguous clauses in smart contracts, making dispute resolution viable for micro-transactions and expanding smart contracts beyond rigid if/then logic.

Vitalik Buterin argues that AI can make formal verification more practical, helping generate specs and proofs to ensure software behaves correctly, potentially transforming critical software development beyond Ethereum.

OpenAI and Paradigm introduce EVMbench, a benchmark for evaluating AI agents' capabilities in detecting, patching, and exploiting smart contract vulnerabilities across 117 curated vulnerabilities from 40 audits. The benchmark demonstrates GPT-5.3-Codex achieving 71% on exploit tasks, significantly outperforming GPT-5's 33.3%, while detection and patching remain more challenging.

Acton is a comprehensive Rust-based CLI toolkit for the full lifecycle of TON smart contract development, including building, testing, and deployment.