RLVE: Scaling Up Reinforcement Learning for Language Models with
  Adaptive Verifiable Environments

Zhiyuan Zeng; Hamish Ivison; Yiping Wang; Lifan Yuan; Shuyue Stella Li; Zhuorui Ye; Siting Li; Jacqueline He; Runlong Zhou; Tong Chen; Chenyang Zhao; Yulia Tsvetkov; Simon Shaolei Du; Natasha Jaques; Hao Peng; Pang Wei Koh; Hannaneh Hajishirzi

Paper · arXiv 2511.07317

RLVE: Scaling Up Reinforcement Learning for Language Models with Adaptive Verifiable Environments

Zhiyuan Zeng, Hamish Ivison, Yiping Wang, Lifan Yuan, Shuyue Stella Li, Zhuorui Ye, Siting Li, Jacqueline He, Runlong Zhou, Tong Chen, Chenyang Zhao, Yulia Tsvetkov, Simon Shaolei Du, Natasha Jaques, Hao Peng, Pang Wei Koh, Hannaneh Hajishirzi

18 upvotesNovember 10, 2025arXiv 预印本

AI 摘要

Reinforcement Learning with Adaptive Verifiable Environments (RLVE) improves language model reasoning by dynamically adjusting problem difficulty, outperforming static environments and traditional RL training.

Reinforcement Learning (RL)Adaptive Verifiable Environments (RLVE)verifiable environmentsprocedural generationalgorithmically verifiable rewardspolicy modelproblem difficulty distributionstatic data distributionsvanishing learning signalsRLVE-Gymenvironment scalinggeneralizable reasoning capabilitiesreasoning benchmarkslanguage models (LMs)

Abstract

We introduce Reinforcement Learning (RL) with Adaptive Verifiable Environments (RLVE), an approach using verifiable environments that procedurally generate problems and provide algorithmically verifiable rewards, to scale up RL for language models (LMs). RLVE enables each verifiable environment to dynamically adapt its problem difficulty distribution to the policy model's capabilities as training progresses. In contrast, static data distributions often lead to vanishing learning signals when problems are either too easy or too hard for the policy. To implement RLVE, we create RLVE-Gym, a large-scale suite of 400 verifiable environments carefully developed through manual environment engineering. Using RLVE-Gym, we show that environment scaling, i.e., expanding the collection of training environments, consistently improves generalizable reasoning capabilities. RLVE with joint training across all 400 environments in RLVE-Gym yields a 3.37% absolute average improvement across six reasoning benchmarks, starting from one of the strongest 1.5B reasoning LMs. By comparison, continuing this LM's original RL training yields only a 0.49% average absolute gain despite using over 3x more compute. We release our code publicly.