Meta-Reinforcement Learning with Self-Reflection for Agentic Search

Open paper

Benchmark	Model	Rank	Results
question-answering-on-natural-questions	MR-Search (Qwen2.5-7B-Base)	#12	EM: 50.2