Mengxuan Hu

Alignment-Weighted DPO: A principled reasoning approach to improve safety alignment

BEACON: Bayesian Optimal Stopping for Efficient LLM Sampling

Improve Temporal Reasoning in Multimodal Large Language Models via Video Contrastive Decoding

Mind Control through Causal Inference: Predicting Clean Images from Poisoned Data

Benign Samples Matter! Fine-tuning On Outlier Benign Samples Severely Breaks Safety

BalancEdit: Dynamically Balancing the Generality-Locality Trade-off in Multi-modal Model Editing

Backdoor in Seconds: Unlocking Vulnerabilities in Large Pre-trained Models via Model Editing

No Free Lunch: Retrieval-Augmented Generation Undermines Fairness in LLMs, Even for Vigilant Users

BalancEdit: Dynamically Balancing the Generality-Locality Trade-off in Multi-modal Model Editing