ProCeedRL: Process Critic with Exploratory Demonstration Reinforcement Learning for LLM Agentic Reasoning
This work addresses a key bottleneck in RL for LLM agentic reasoning, offering a novel method to enhance exploration in long-horizon tasks, though it appears incremental in its approach to improving existing RL techniques.
The paper tackles the challenge of applying reinforcement learning to multi-turn agentic tasks with large language models by addressing a structural failure mode where suboptimal actions lead to cumulative errors, proposing ProCeedRL to actively intervene and guide exploration, resulting in significant improvements in exploration efficiency and performance on complex tasks.
Reinforcement Learning (RL) significantly enhances the reasoning abilities of large language models (LLMs), yet applying it to multi-turn agentic tasks remains challenging due to the long-horizon nature of interactions and the stochasticity of environmental feedback. We identify a structural failure mode in agentic exploration: suboptimal actions elicit noisy observations into misleading contexts, which further weaken subsequent decision-making, making recovery increasingly difficult. This cumulative feedback loop of errors renders standard exploration strategies ineffective and susceptible to the model's reasoning and the environment's randomness. To mitigate this issue, we propose ProCeedRL: Process Critic with Explorative Demonstration RL, shifting exploration from passive selection to active intervention. ProCeedRL employs a process-level critic to monitor interactions in real time, incorporating reflection-based demonstrations to guide agents in stopping the accumulation of errors. We find that this approach significantly exceeds the model's saturated exploration performance, demonstrating substantial exploratory benefits. By learning from exploratory demonstrations and on-policy samples, ProCeedRL significantly improves exploration efficiency and achieves superior performance on complex deep search and embodied tasks.