PPL
Jiatan Huang
Papers (4)
Where LLM Annotators Fail: Label-Free Learning on Graphs with LLMsIn-Context Optimization for Retrieval-Augmented Generation: A Gradient-Descent PerspectiveRICE-PO: Turning Retrieval Interactions into Credit Signals for Reasoning AgentsSame Signal, Opposite Meaning: Direction-Informed Adaptive Learning for LLM Agents