Selected work

Representative Papers

ChainClaw: A Layered Agent Framework for Reliable On-Chain Execution

Aug 06, 2026

This work addresses three core challenges faced by general-purpose large language model agents in blockchain environments: insufficient reactivity, irreversible operations, and unobservable system states. To overcome these limitations, the authors propose ChainClaw, the first hierarchical agent framework specifically designed for blockchain settings. ChainClaw integrates an event-driven orchestration layer, simulation-based safe reasoning, and on-chain monitoring within a unified runtime architecture, augmented by a cross-layer memory mechanism to ensure coherent decision-making. The framework establishes a secure execution pipeline through key components including event ingestion, transaction simulation, action safeguarding, on-chain read adapters, and transaction monitoring. Evaluated across seven tasks spanning four distinct categories, ChainClaw demonstrates substantial improvements over existing approaches in both safety and task completion rates.

0 citationsRead paper

Towards Compositional Generalization in LLMs for Smart Contract Security: A Case Study on Reentrancy Vulnerabilities

Jan 11, 2026arXiv.org

This work addresses the limited generalization capability of large language models (LLMs) in detecting reentrancy vulnerabilities in smart contracts, where they often underperform compared to traditional static analysis tools. To overcome this limitation, the authors propose a post-training approach based on atomic task decomposition and fusion. Specifically, reentrancy detection is decomposed into four linearly independent atomic tasks, and the model is fine-tuned using synthetic data, LoRA-based low-rank adapters, and structural information from control-flow and data-flow graphs extracted by Slither. A low-rank normalization fusion strategy is further introduced to enhance compositional generalization. This method achieves 98.2% accuracy in reentrancy detection and demonstrates a 20% higher recall than conventional static analysis tools across 31 real-world smart contracts, marking the first successful application of compositional generalization with LLMs in smart contract security.

0 citationsRead paper

Recent publications

Latest Papers

ChainClaw: A Layered Agent Framework for Reliable On-Chain Execution

Aug 06, 2026

This work addresses three core challenges faced by general-purpose large language model agents in blockchain environments: insufficient reactivity, irreversible operations, and unobservable system states. To overcome these limitations, the authors propose ChainClaw, the first hierarchical agent framework specifically designed for blockchain settings. ChainClaw integrates an event-driven orchestration layer, simulation-based safe reasoning, and on-chain monitoring within a unified runtime architecture, augmented by a cross-layer memory mechanism to ensure coherent decision-making. The framework establishes a secure execution pipeline through key components including event ingestion, transaction simulation, action safeguarding, on-chain read adapters, and transaction monitoring. Evaluated across seven tasks spanning four distinct categories, ChainClaw demonstrates substantial improvements over existing approaches in both safety and task completion rates.

0 citationsRead paper

Towards Compositional Generalization in LLMs for Smart Contract Security: A Case Study on Reentrancy Vulnerabilities

Jan 11, 2026arXiv.org

This work addresses the limited generalization capability of large language models (LLMs) in detecting reentrancy vulnerabilities in smart contracts, where they often underperform compared to traditional static analysis tools. To overcome this limitation, the authors propose a post-training approach based on atomic task decomposition and fusion. Specifically, reentrancy detection is decomposed into four linearly independent atomic tasks, and the model is fine-tuned using synthetic data, LoRA-based low-rank adapters, and structural information from control-flow and data-flow graphs extracted by Slither. A low-rank normalization fusion strategy is further introduced to enhance compositional generalization. This method achieves 98.2% accuracy in reentrancy detection and demonstrates a 20% higher recall than conventional static analysis tools across 31 real-world smart contracts, marking the first successful application of compositional generalization with LLMs in smart contract security.

0 citationsRead paper