Tsung-Yi Ho

ReAlign: Safety-Aligning Reasoning Models with Verifier-Guided Reinforcement Learning

Attention Slipping: A Mechanistic Understanding of Jailbreak Attacks and Defenses in LLMs

PermLLM: Learnable Channel Permutation for N:M Sparse Large Language Models

CoP: Agentic Red-teaming for Large Language Models using Composition of Principles

CARE: Decoding-Time Safety Alignment via Rollback and Introspection Intervention

Test Time Augmentations are Worth One Million Images for Out-of-Distribution Detection

Your Task May Vary: A Systematic Understanding of Alignment and Safety Degradation when Fine-tuning LLMs

Defensive Prompt Patch: A Robust and Generalizable Defense of Large Language Models against Jailbreak Attacks

Elephant in the Room: Unveiling the Pitfalls of Human Proxies in Alignment

GRE Score: Generative Risk Evaluation for Large Language Models