让大模型找回训练中遗忘的推理能力，无需重训。
Temporal Sampling for Forgotten Reasoning in LLMs
- 从训练轨迹多个检查点采样输出，恢复遗忘的解法。
- 在多个基准上提升4~19分，且多数情况性能稳定增长。
- 仅存适配器权重即可实现，适合资源受限场景使用。
微调大语言模型旨在提升其推理能力，但我们发现一种反直觉现象：模型在训练过程中常忘记先前能正确解答的问题。我们称此为时间遗忘，且该现象普遍存在于不同模型规模、微调方法（强化学习与监督微调）及多个推理基准中。为此，我们提出时序采样（Temporal Sampling），一种简单的解码策略，从训练轨迹的多个检查点中抽取输出。该方法无需重训或集成即可恢复遗忘的解法，在多个基准上带来4至19分的Pass@k提升，并在Majority@k上实现一致增益。我们进一步将方法扩展至LoRA适配模型，证明仅存储检查点间的适配器权重即可获得类似收益，存储开销极小。通过利用训练过程中的时序多样性，时序采样提供了一种高效、实用的方法，以挖掘隐藏的推理能力，并重新思考如何评估大模型。
原文摘要 · Abstract (English)
Fine-tuning large language models (LLMs) is intended to improve their reasoning capabilities, yet we uncover a counterintuitive effect: models often forget how to solve problems they previously answered correctly during training. We term this phenomenon temporal forgetting and show that it is widespread across model sizes, fine-tuning methods (both Reinforcement Learning and Supervised Fine-Tuning), and multiple reasoning benchmarks. To address this gap, we introduce Temporal Sampling, a simple decoding strategy that draws outputs from multiple checkpoints along the training trajectory. This approach recovers forgotten solutions without retraining or ensembling, and leads to substantial improvements in reasoning performance, gains from 4 to 19 points in Pass@k and consistent gains in Majority@k across several benchmarks. We further extend our method to LoRA-adapted models, demonstrating that storing only adapter weights across checkpoints achieves similar benefits with minimal storage cost. By leveraging the temporal diversity inherent in training, Temporal Sampling offers a practical, compute-efficient way to surface hidden reasoning ability and rethink how we evaluate LLMs.
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。