J6: Jacobian-Driven Role Attribution for Multi-Objective Prompt Optimization in LLMs

Yao Wu

In large language model (LLM) adaptation, balancing multiple optimization objectives such as improving factuality (heat) and increasing confidence (via low entropy) poses a fundame…