Search-R-1

The 'Search-R-1' dataset/benchmark is used to evaluate the performance of LLM-based search agents in multi-turn interactions by assessing their ability to manage memory and maintain relevant context.

Papers using Search-R-1 (3)

MemSearcher: Training LLMs to Reason, Search and Manage Memory via End-to-End Reinforcement Learning2025RLFactory: A Plug-and-Play Reinforcement Learning Post-Training Framework for LLM Multi-Turn Tool-Use2025Agent-as-Tool: A Study on the Hierarchical Decision Making with Reinforcement Learning2025