Shorten After You’re Right: Lazy Length Penalties for Reasoning RL

ACL ARR 2026 January Submission8971 Authors

Keywords: llm reasoning, reinforcement learning, long to short reasoning

Abstract: Long-reasoning models achieve strong accuracy on complex reasoning tasks, but their extended reasoning trajectories incur substantial memory and latency costs. Several existing shortening methods rely on additional supervision or multi-stage post-training, which primarily reduces inference length and does not reduce the rollout tokens during on-policy reinforcement learning (RL). We instead target on-policy response shortening, aiming to improve both inference efficiency and RL training throughput. However, because on-policy RL couples optimization with exploration, naively penalizing length can destabilize training and suppress exploration. To impose length pressure safely, we propose a lazy length penalty integrated into the rule-based RL pipeline: it activates only on correct trajectories, only after training accuracy enters a stably improving regime, and only when responses exceed a tolerance band beyond the minimal correct length. Across four settings, our method significantly reduces response length without extra training stages while maintaining or improving performance. In a logic reasoning setting, we achieve a 40% reduction in step-averaged response length alongside a 14-point gain in performance. For math problems, we reduce step-averaged response length by 33% while preserving performance.

Paper Type: Long

Research Area: Language Models

Research Area Keywords: chain-of-thought

Contribution Types: NLP engineering experiment

Languages Studied: English

Submission Number: 8971