What Does Multi-Turn Reinforcement Learning Really Improve in LLM Agents?

Multi-turn agentic reinforcement learning (RL) improves long-horizon task performance, but those gains are highly sensitive to interaction constraints, according to an evaluation using Pass@(K,T) across sampling breadth K and interaction depth T. The evaluation shows trajectory-level RL advantages depend on sufficient interaction depth, indicating that reported gains for LLM agents hinge on how many turns and samples the agent is allowed.

While multi-turn agentic reinforcement learning RL excels on long-horizon tasks, its performance gains remain highly sensitive to interaction constraints. Evaluating across sampling breadth K and interaction depth T via Pass@ K,T , we show that trajectory-level RL advantages depend on sufficient t