Improving LLM Reasoning through Interpretable Role-Playing Steering

LLMSVLRMRe-assign community