research
RAPO framework improves LLM agent reasoning by combining retrieval with reinforcement learning
Researchers introduce RAPO (Retrieval-Augmented Policy Optimization), a reinforcement learning framework that improves LLM agent reasoning by incorporating off-policy retrieval signals during training. The method achieves an average 5.0% performance gain across fourteen datasets and delivers 1.2x faster training efficiency compared to existing agentic RL approaches.