重复训练比扩充数据更有效，能显著提升推理模型性能。
Data Repetition Beats Data Scaling in Long-CoT Supervised Fine-Tuning
- 用小数据多轮重复训练，比大数据单轮训练效果更好。
- 400样本训练128轮在AIME/GPQA上领先51200样本1轮12-26个百分点。
- 以词元准确率作为停止信号，可避免盲目扩数据，适合资源有限者。
链式思维（Chain-of-Thought）数据上的监督微调（SFT）是提升推理型语言模型的关键步骤。传统机器学习认为更多唯一样本能提升泛化能力，但反直觉的是，我们发现重复训练更优：在固定更新预算下，小数据集多轮训练优于大数据集单轮训练。在AIME'24/25和GPQA基准上，Olmo3-7B模型在400个样本上训练128轮，相比51200样本训练1轮，性能提升12-26个百分点，且无额外灾难性遗忘。我们发现词元准确率能可靠指示重复饱和点：当模型完全记忆后，进一步增加训练轮次性能不再提升，该规律在所有设置中一致。这些结果为推理型SFT提供了实用方法——以词元准确率为停止标准，用扩展训练轮次替代昂贵的无向数据扩展。我们提出‘重复优势’现象：完全记忆时反而泛化更好，这成为理解大模型训练动态的新开放问题。
原文摘要 · Abstract (English)
Supervised fine-tuning (SFT) on chain-of-thought data is an essential post-training step for reasoning language models. Standard machine learning intuition suggests that training with more unique training samples yields better generalization. Counterintuitively, we show that SFT benefits from repetition: under a fixed update budget, training for more epochs on smaller datasets outperforms single-epoch training on larger datasets. On AIME'24/25 and GPQA benchmarks, Olmo3-7B trained for 128 epochs on 400 samples outperforms the equivalent 1 epoch on 51200 samples by 12-26 percentage points, with no additional catastrophic forgetting. We find that training token accuracy reliably signals when repetition has saturated; improvements from additional epochs plateau at full memorization, a pattern consistent across all settings. These findings provide a practical approach for reasoning SFT, where scaling epochs with token accuracy as a stopping criterion can replace expensive undirected data scaling. We pose the repetition advantage, where full memorization coincides with improved generalization, as a new open problem for the community in understanding the training dynamics of large language models.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。