Decoupled Smart Contract Audits: Lightweight LLM Framework via Distillation and Aggregation
Researchers have developed a new framework for auditing smart contracts using lightweight Large Language Models. This system decouples the auditing process into four stages: detection, explanation, severity classification, and remediation. By employing techniques like knowledge distillation and a custom aggregation strategy, the framework achieves high accuracy in vulnerability detection and generative explanation tasks, outperforming larger open-source models. AI
IMPACT This research introduces a more efficient method for LLM-based smart contract auditing, potentially improving security in decentralized applications.