Contrastive Reflection for Iterative Prompt Optimization

arXiv:2606.30840v1 Announce Type: new Abstract: LLM agents are becoming central to information retrieval: they issue retrieval queries, synthesize answers, and increasingly serve as judges for IR evaluation.