Junyuan Hong

LLMs Can Get "Brain Rot"!

LLM Unlearning via Calibrated and Tokenized Negative Preference Alignment

POPS: Recovering Unlearned Multi-Modality Knowledge in MLLMs with Fine-tuning and Prompt-based Attacks

LoX: Low-Rank Extrapolation Robustifies LLM Safety Against Fine-tuning

SEAL: Steerable Reasoning Calibration of Large Language Models for Free

More is Less: The Pitfalls of Multi-Model Synthetic Preference Data in DPO Safety Alignment

GuardAgent: Safeguard LLM Agents via Knowledge-Enabled Reasoning

GuardAgent: Safeguard LLM Agent by a Guard Agent via Knowledge-Enabled Reasoning