Lookahead Tree-Based Rollouts for Enhanced Trajectory-Level Exploration in Reinforcement Learning with Verifiable Rewards

TL;DR

This paper introduces Lookahead Tree-Based Rollouts (LATR), a novel strategy to increase trajectory diversity in reinforcement learning, leading to faster policy learning and improved reasoning performance in language models.

Contribution

LATR explicitly promotes trajectory diversity through branching and lookahead, significantly enhancing policy learning efficiency and effectiveness in RL with verifiable rewards.

Findings

LATR accelerates policy learning by 131% on average.

LATR improves final pass@1 performance by 4.2%.

LATR outperforms stochastic sampling in diverse reasoning tasks.

Abstract

Reinforcement Learning with Verifiable Rewards (RLVR), particularly with algorithms like Group Relative Policy Optimization (GRPO), has proven highly effective in enhancing the reasoning capabilities of large language models. However, a critical bottleneck in current pipelines lies in the limited diversity of sampled trajectories during group rollouts. Homogeneous trajectories and their associated rewards would diminish the return signals for policy updates, thereby hindering effective policy learning. This lack of diversity stems primarily from token-level stochastic sampling, where local variations are likely to collapse into near-identical reasoning paths. To address this limitation, we propose Lookahead Tree-Based Rollouts (LATR), a novel rollout strategy designed to explicitly promotes trajectory-level diversity by enforcing branching into different candidate tokens likely to yield distinct continuations. Specifically, LATR iteratively operates in three stages: (1) branching at high-uncertainty generation steps, (2) performing lookahead simulation for each new branch, and (3) pruning branches that exhibits prolonged similarity during simulation. Compared with stochastic Sampling, LATR accelerates policy learning by 131% on average and improves final pass@1 performance by 4.2% on both GRPO and Dynamic sAmpling Policy Optimization (DAPO) algorithms across different reasoning tasks. Our code and data are publicly available at https://github.com/starreeze/latr.