OctoThinker: Mid-training Incentivizes Reinforcement Learning Scaling

• URL: http://arxiv.org/abs/2506.20512v1
• Date: Wed, 25 Jun 2025 14:58:13 GMT
• Title: OctoThinker: Mid-training Incentivizes Reinforcement Learning Scaling
• Authors: Zengzhi Wang, Fan Zhou, Xuefeng Li, Pengfei Liu,
• Abstract summary: Different base language model families, such as Llama and Qwen, exhibit divergent behaviors during post-training with reinforcement learning (RL)<n>Our study reveals that high-quality mathematical corpora, such as MegaMath-Web-Pro, significantly improve both base model and RL performance.<n>We introduce a two-stage mid-training strategy, Stable-then-Decay, in which base models are first trained on 200B tokens with a constant learning rate, followed by 20B tokens across three CoT-focused branches with learning rate decay.
• Score: 29.818409458662344
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Different base language model families, such as Llama and Qwen, exhibit divergent behaviors during post-training with reinforcement learning (RL), especially on reasoning-intensive tasks. What makes a base language model suitable for reinforcement learning? Gaining deeper insight into this question is essential for developing RL-scalable foundation models of the next generation. In this work, we investigate how mid-training strategies shape RL dynamics, focusing on two representative model families: Qwen and Llama. Our study reveals that (1) high-quality mathematical corpora, such as MegaMath-Web-Pro, significantly improve both base model and RL performance, while existing alternatives (e.g., FineMath-4plus) fail to do so; (2) further adding QA-style data, particularly long chain-of-thought (CoT) reasoning examples, enhances RL outcomes, and instruction data further unlocks this effect; (3) while long-CoT improves reasoning depth, it can also induce verbosity of model responses and unstability of RL training, underscoring the importance of data formatting; (4) scaling mid-training consistently leads to stronger downstream RL performance. Building on these insights, we introduce a two-stage mid-training strategy, Stable-then-Decay, in which base models are first trained on 200B tokens with a constant learning rate, followed by 20B tokens across three CoT-focused branches with learning rate decay. This yields OctoThinker, a family of models demonstrating strong RL compatibility and closing the performance gap with more RL-friendly model families, i.e., Qwen. We hope our work will help shape pre-training strategies for foundation models in the RL era. To support further research, we release our open-source models along with a curated math reasoning-intensive corpus of over 70 billion tokens (i.e., MegaMath-Web-Pro-Max).

Related papers

• Learning What Reinforcement Learning Can't: Interleaved Online Fine-Tuning for Hardest Questions [28.962415274754537]
  Large language model (LLM) reasoning has shown that sophisticated behaviors such as planning and self-reflection can emerge through reinforcement learning (RL)<n>We introduce a novel training approach, textbfReLIFT (textbfReinforcement textbfL textbfInterleaved with Online textbfFine-textbfTuning)<n>In ReLIFT, the model is primarily trained using RL, but when it encounters challenging questions, high-quality solutions are collected for fine-tuning, and the training process alternate
  arXiv  Detail & Related papers  (2025-06-09T08:11:20Z)
• Beyond Accuracy: Dissecting Mathematical Reasoning for LLMs Under Reinforcement Learning [82.43575191712726]
  We introduce a fine-grained analytic framework to dissect the impact ofReinforcement learning on reasoning.<n>Our framework specifically investigates key elements that have been hypothesized to benefit from RL training.
  arXiv  Detail & Related papers  (2025-06-05T07:53:59Z)
• ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models [89.37819814048288]
  We introduce ProRL, a novel training methodology that incorporates KL divergence control, reference policy, and a diverse suite of tasks.<n>Our empirical analysis reveals that RL-trained models consistently outperform base resetting models across a wide range of pass@k evaluations.<n>These findings offer new insights into the conditions under which RL meaningfully expands reasoning boundaries in language models.
  arXiv  Detail & Related papers  (2025-05-30T17:59:01Z)
• Advancing Multimodal Reasoning via Reinforcement Learning with Cold Start [24.244577648817188]
  "aha moment" patterns are often attributed to emergent properties from reinforcement learning (RL)<n>We present a comprehensive study on enhancing multimodal reasoning through a two-stage approach.<n>Our experiments show that this combined approach consistently outperforms both SFT-only and RL-only methods.
  arXiv  Detail & Related papers  (2025-05-28T13:21:38Z)
• AceReason-Nemotron: Advancing Math and Code Reasoning through Reinforcement Learning [50.02117478165099]
  We show that large-scale reinforcement learning can significantly enhance the reasoning capabilities of strong, small- and mid-sized models.<n>We propose a simple yet effective approach: first training on math-only prompts, then on code-only prompts.
  arXiv  Detail & Related papers  (2025-05-22T08:50:47Z)
• Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model? [67.30809748319486]
  Reinforcement Learning with Verifiable Rewards (RLVR) has recently demonstrated notable success in enhancing the reasoning performance of large language models (LLMs)<n>This study critically examines the current state of RLVR.<n>We find that the current training setup does not elicit fundamentally new reasoning patterns.
  arXiv  Detail & Related papers  (2025-04-18T17:59:56Z)
• Echo Chamber: RL Post-training Amplifies Behaviors Learned in Pretraining [74.83412846804977]
  Reinforcement learning (RL)-based fine-tuning has become a crucial step in post-training language models.<n>We present a systematic end-to-end study of RL fine-tuning for mathematical reasoning by training models entirely from scratch.
  arXiv  Detail & Related papers  (2025-04-10T17:15:53Z)
• SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wild [46.25416990387885]
  Long chain-of-thought (CoT) reasoning can naturally emerge through a simple reinforcement learning framework with rule-based rewards.<n>We investigate zero RL training across 10 diverse base models including LLama3-8B, Mistral-7B/24B, DeepSeek-Math-7B, Qwen2.5-math-7B, and all Qwen2.5 models from 0.5B to 32B.
  arXiv  Detail & Related papers  (2025-03-24T17:06:10Z)
• An Empirical Study on Eliciting and Improving R1-like Reasoning Models [90.52239241349504]
  scaling RL training has become a central technique for implementing such reasoning models.<n>We demonstrate that our RL training approach consistently improves the Qwen2.5-32B base models.<n>We also explore the use of tool manipulation, finding that it significantly boosts the reasoning performance of large reasoning models.
  arXiv  Detail & Related papers  (2025-03-06T15:34:27Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.