SAAS: Self-Aware Reinforcement Learning for Over-Search Mitigation in Agentic Search

Researchers introduced SAAS, a reinforcement learning framework for agentic search that helps LLMs decide when to stop searching.

SAAS models search boundaries, uses boundary-aware rewards, and trains in stages to reduce redundant retrieval.

The approach targets over-search in multi-hop question answering, lowering cost and inference latency without harming answer quality.