7 papers
From Data to Behavior: Predicting Unintended Model Behaviors Before Training
Mengru Wang, Zhenqian Xu, Junfeng Fang +4
Large Language Models (LLMs) can acquire unintended biases from seemingly benign training data even without explicit cues or malicious content. Existing methods struggle to detect…
Automating Steering for Safe Multimodal Large Language Models
Lyucheng Wu, Mengru Wang, Ziwen Xu +4
Recent progress in Multimodal Large Language Models (MLLMs) has unlocked powerful cross-modal reasoning abilities, but also raised new safety concerns, particularly when faced with…
Beyond Prompt Engineering: Robust Behavior Control in LLMs via Steering Target Atoms
Mengru Wang, Ziwen Xu, Shengyu Mao +4
Precise control over language model generation is vital for ensuring both safety and reliability. Although prompt engineering and steering are commonly used to intervene in model b…
Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training
Mengru Wang, Xingyu Chen, Yue Wang +12
Mixture-of-Experts (MoE) architectures within Large Reasoning Models (LRMs) have achieved impressive reasoning capabilities by selectively activating experts to facilitate structur…
EasyEdit2: An Easy-to-use Steering Framework for Editing Large Language Models
Ziwen Xu, Shuxun Wang, Kewei Xu +7
In this paper, we introduce EasyEdit2, a framework designed to enable plug-and-play adjustability for controlling Large Language Model (LLM) behaviors. EasyEdit2 supports a wide ra…
LookAhead Tuning: Safer Language Models via Partial Answer Previews
Kangwei Liu, Mengru Wang, Yujie Luo +7
Fine-tuning enables large language models (LLMs) to adapt to specific domains, but often compromises their previously established safety alignment. To mitigate the degradation of m…