9 papers
SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment
Hao Li, Jingkun An, Zijun Song +8
Aligning Large Language Models (LLMs) with human values often degrades their general capabilities, termed the alignment tax. Existing methods mitigate this by balancing dual object…
StepCodeReasoner: Aligning Code Reasoning with Stepwise Execution Traces via Reinforcement Learning
Hao Wang, Rui Li, Lei Sha +1
Existing code reasoning methods primarily supervise final code outputs, ignoring intermediate states, often leading to reward hacking where correct answers are obtained through inc…
Be Your Own Red Teamer: Safety Alignment via Self-Play and Reflective Experience Replay
Hao Wang, Yanting Wang, Hao Li +2
Large Language Models (LLMs) have achieved remarkable capabilities but remain vulnerable to adversarial ``jailbreak'' attacks designed to bypass safety guardrails. Current safety a…
Layer-Aware Representation Filtering: Purifying Finetuning Data to Preserve LLM Safety Alignment
Hao Li, Lijun Li, Zhenghao Lu +4
With rapid advancement and increasing accessibility of LLMs, fine-tuning aligned models has become a critical step for adapting them to real-world applications, which makes the saf…
HauntAttack: When Attack Follows Reasoning as a Shadow
Jingyuan Ma, Rui Li, Zheng Li +4
Emerging Large Reasoning Models (LRMs) consistently excel in mathematical and reasoning tasks, showcasing remarkable capabilities. However, the enhancement of reasoning abilities a…
Towards Harmonized Uncertainty Estimation for Large Language Models
Rui Li, Jing Long, Muge Qi +4
To facilitate robust and trustworthy deployment of large language models (LLMs), it is essential to quantify the reliability of their generations through uncertainty estimation. Wh…
Be a Multitude to Itself: A Prompt Evolution Framework for Red Teaming
Rui Li, Peiyi Wang, Jingyuan Ma +3
Large Language Models (LLMs) have gained increasing attention for their remarkable capacity, alongside concerns about safety arising from their potential to produce harmful content…
How Far are LLMs from Being Our Digital Twins? A Benchmark for Persona-Based Behavior Chain Simulation
Rui Li, Heming Xia, Xinfeng Yuan +4
Recently, LLMs have garnered increasing attention across academic disciplines for their potential as human digital twins, virtual proxies designed to replicate individuals and auto…
Plug-and-Play Training Framework for Preference Optimization
Jingyuan Ma, Rui Li, Zheng Li +2
Recently, preference optimization methods such as DPO have significantly enhanced large language models (LLMs) in wide tasks including dialogue and question-answering. However, cur…