Prune as You Generate: Online Rollout Pruning for Faster and Better RLVR

TL;DR

This paper introduces arrol, an online rollout pruning method for RLVR that improves training efficiency and accuracy by early pruning of rollouts during generation using a learned quality predictor.

Contribution

The paper proposes arrol, a novel online rollout pruning technique that enhances RLVR by dynamically pruning rollouts and training a lightweight predictor for success probability.

Findings

Achieves up to 1.7x training speedup.

Improves average accuracy by +2.30 to +2.99.

Yields up to +8.33 gains in test-time scaling.

Abstract

Reinforcement Learning with Verifiable Rewards (RLVR) has significantly advanced the reasoning capabilities of Large Language Models (LLMs). However, methods such as GRPO and DAPO suffer from substantial computational cost, since they rely on sampling many rollouts for each prompt. Moreover, in RLVR the relative advantage is often sparse: many samples become nearly all-correct or all-incorrect, yielding low within-group reward variance and thus weak learning signals. In this paper, we introduce arrol (Accelerating RLVR via online Rollout Pruning), an online rollout pruning method that prunes rollouts during generation while explicitly steering the surviving ones more correctness-balanced to enhance learning signals. Specifically, arrol trains a lightweight quality head on-the-fly to predict the success probability of partial rollouts and uses it to make early pruning decisions. The learned quality head can further weigh candidates to improve inference accuracy during test-time scaling. To improve efficiency, we present a system design that prunes rollouts inside the inference engine and re-batches the remaining ones for log-probability computation and policy updates. Across GRPO and DAPO on Qwen-3 and LLaMA-3.2 models (1B-8B), arrol improves average accuracy by +2.30 to +2.99 while achieving up to 1.7x training speedup, and yielding up to +8.33 additional gains in average accuracy in test-time scaling. The code is available at https://github.com/Hsu1023/ARRoL.