Yue Xu

EvoPoC introduces a knowledge-driven agentic system that automates the synthesis of verifiable and economically viable exploits for DeFi smart contracts, achieving high recall and significant revenue recovery rates.

This paper introduces Security Cube, a comprehensive, multi-dimensional framework for evaluating LLM robustness against jailbreak attacks, providing a systematic taxonomy and benchmark analysis of existing attacks and defenses.

The paper proposes EVA, a novel framework that uses direct model editing to surgically correct specific neurons responsible for jailbreaking vulnerabilities in LLMs and VLMs, achieving robust safety alignment without performance degradation.