This is the official code for the paper "Booster: Tackling Harmful Fine-tuning for Large Language Models via Attenuating Harmful Perturbation" (ICLR2025 Oral).

GTLLMZoo: A comprehensive framework that aggregates LLM benchmark data from multiple sources with an interactive UI for efficient model comparison, filtering, and evaluation across performance, safety, and efficiency metrics.

This is the official code for the paper "Safety Tax: Safety Alignment Makes Your Large Reasoning Models Less Reasonable".

A survey on harmful fine-tuning attack for large language model

A Survey on Large Language Model-Based Game Agents

This is the official code for the paper "Virus: Harmful Fine-tuning Attack for Large Language Models Bypassing Guardrail Moderation"

This is the official code for the paper "Vaccine: Perturbation-aware Alignment for Large Language Models" (NeurIPS2024)