4 papers
UniSteer: Text-Guided Flow Matching in Activation Space for Versatile LLM Steering
Yingdong Shi, Ruiming Zhang, Changming Li +4
Activation-based control steers large language models (LLMs) by intervening on their internal representations during inference, and has emerged as an effective paradigm for control…
Interpreting and Controlling LLM Reasoning through Integrated Policy Gradient
Changming Li, Kaixing Zhang, Haoyun Xu +4
Large language models (LLMs) demonstrate strong reasoning abilities in solving complex real-world problems. Yet, the internal mechanisms driving these complex reasoning behaviors r…
Discovering Influential Neuron Path in Vision Transformers
Yifan Wang, Yifei Liu, Yingdong Shi +5
Vision Transformer models exhibit immense power yet remain opaque to human understanding, posing challenges and risks for practical applications. While prior research has attempted…
Dissecting and Mitigating Diffusion Bias via Mechanistic Interpretability
Yingdong Shi, Changming Li, Yifan Wang +5
Diffusion models have demonstrated impressive capabilities in synthesizing diverse content. However, despite their high-quality outputs, these models often perpetuate social biases…