Deep Dense Exploration for LLM Reinforcement Learning via Pivot-Driven Resampling
This addresses a key bottleneck in LLM reinforcement learning for tasks like mathematical reasoning, though it appears incremental as it builds on prior methods like GRPO.
The paper tackles the challenge of effective exploration in reinforcement learning for large language models by proposing Deep Dense Exploration (DDE), which focuses on deep, recoverable states within unsuccessful trajectories; experiments on mathematical reasoning benchmarks show it consistently outperforms existing methods like GRPO and tree-based approaches.
Effective exploration is a key challenge in reinforcement learning for large language models: discovering high-quality trajectories within a limited sampling budget from the vast natural language sequence space. Existing methods face notable limitations: GRPO samples exclusively from the root, saturating high-probability trajectories while leaving deep, error-prone states under-explored. Tree-based methods blindly disperse budgets across trivial or unrecoverable states, causing sampling dilution that fails to uncover rare correct suffixes and destabilizes local baselines. To address this, we propose Deep Dense Exploration (DDE), a strategy that focuses exploration on $\textit{pivots}$-deep, recoverable states within unsuccessful trajectories. We instantiate DDE with DEEP-GRPO, which introduces three key innovations: (1) a lightweight data-driven utility function that automatically balances recoverability and depth bias to identify pivot states; (2) local dense resampling at each pivot to increase the probability of discovering correct subsequent trajectories; and (3) a dual-stream optimization objective that decouples global policy learning from local corrective updates. Experiments on mathematical reasoning benchmarks demonstrate that our method consistently outperforms GRPO, tree-based methods, and other strong baselines.