Publications: - PATCHAGENT: A Practical Program Repair Agent Mimicking Human Expertise (USENIX Security 2025, CSAW 2025 Finalist) - GPO: Learning from Critical Steps to Improve LLM Reasoning (NIPS 2025, Featured in MIT Technology Review China) - Soft-Label Integration for Robust Toxicity Classification (NIPS 2024, Featured in MIT Technology Review China) - GPTFuzzer: Red Teaming Large Language Models with Auto-Generated Jailbreak Prompts (arXiv 2023, Geekcon 2023 Annual Themed Debate Breakthrough Awards) - StateMask: Explaining Deep Reinforcement Learning through State Mask (NIPS 2023) Awards: - Geekcon 2023 Annual Themed Debate Breakthrough Awards - GPTFuzzer project - CSAW 2025 Finalist - PATCHAGENT project
Research Experience
Involved in multiple research projects, including PATCHAGENT, GPO, BlockScan, etc.
Education
B.S. degree: Shanghai Jiao Tong University (2021). Ph.D. degree: Northwestern University (Advisor: Prof. Xinyu).
Background
Research Interests: Large Language Models and cybersecurity. Professional Field: Computer Science. Background: Jiahao Yu is a last-year computer science Ph.D. candidate at Northwestern University, currently on the faculty job market.