NumScout is a novel approach combining Large Language Models with symbolic execution to detect numerical defects in high-value smart contracts.

• Identifies complex numerical vulnerabilities in mathematical financial models used by DApps
• Uses an innovative LLM-pruning technique to enhance symbolic execution efficiency
• Focuses on detecting flaws that could lead to financial losses in high-TVL contracts
• Addresses a critical security gap in traditional smart contract auditing

This research is particularly valuable for blockchain security teams as it targets sophisticated mathematical errors that standard tools miss, potentially preventing significant financial damage in decentralized finance applications.

NumScout: Unveiling Numerical Defects in Smart Contracts using LLM-Pruning Symbolic Execution