{"ID":2863368,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2509.24375","arxiv_id":"2509.24375","title":"Reinforcement Mid-Training","abstract":"The development of state-of-the-art large language models is commonly understood as a two-stage process involving pre-training and post-training. We point out the need for an additional intermediate stage called reinforcement mid-training with potential for strong performance gains. In this paper, we formally define the problem and identify three key challenges: (1) inefficient training due to excessive reasoning steps, (2) disregard of the imbalanced token entropy distribution, and (3) underutilization of token information. To address these challenges, we propose RMT, a framework for efficient, adaptive, and unified reinforcement mid-training with various innovative components. In particular, we first introduce a dynamic token budget mechanism that constrains unnecessary reasoning steps and mitigates model overthinking. Next, we design a curriculum-based adaptive sampling method that fosters a progressive learning trajectory from easy to hard tokens. Finally, we present a dual training strategy that combines reinforcement learning with next-token prediction, ensuring targeted learning on key tokens and full exploitation of all token information. Extensive experiments demonstrate the superiority of RMT over state-of-the-art methods, achieving up to +64.91% performance improvement with only 21% of the reasoning length in language modeling. We also show that checkpoints obtained after reinforcement mid-training can benefit the subsequent post-training, yielding up to +18.76% improvement in the mathematical domain.","short_abstract":"The development of state-of-the-art large language models is commonly understood as a two-stage process involving pre-training and post-training. We point out the need for an additional intermediate stage called reinforcement mid-training with potential for strong performance gains. In this paper, we formally define th...","url_abs":"https://arxiv.org/abs/2509.24375","url_pdf":"https://arxiv.org/pdf/2509.24375v1","authors":"[\"Yijun Tian\",\"Shaoyu Chen\",\"Zhichao Xu\",\"Yawei Wang\",\"Jinhe Bi\",\"Peng Han\",\"Wei Wang\"]","published":"2025-09-29T07:21:24Z","proceeding":"cs.CL","tasks":"[\"cs.CL\"]","methods":"[\"Reinforcement Learning\",\"Language Model\"]","has_code":false}