发表机构
MetaCircle; Tsinghua University; Peking University; Shanghai Qizhi Institute（MetaCircle; 清华大学; 北京大学; 上海期智研究院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文研究数据重放下模型出现的分叉过拟合现象，揭示其机制源于n-gram模块放大重复更新效应，并指出这是自动研究代理技巧的意外副产品，需谨慎对待其结果。
AI 中文摘要
本文研究了分叉（forking）现象，这是NanoGPT自动研究（autoresearch）中发现的一种泛化失败。在数据重放（data replay）下，具有过编码n-gram记忆分支的模型在epoch边界处表现出训练损失和验证损失的急剧分离，形似叉子。我们在受控的原始NanoGPT设置中研究这一现象，并在具有Engram的DeepSeek风格模型中复现了它。机制上，重复更新强化了训练中观察到的延续（continuations），同时抑制了未见延续的概率，其损失随每次遍历而增长。n-gram模块创建了弱交互的上下文特定子空间，放大了这一效应。低频上下文贡献了大部分差距，而更大的训练预算和高度拥挤的表则抑制了它。我们还在短预算、高度重复的SFT和类似RL的场景中观察到分叉。本文的贡献有两方面：（1）分叉揭示了深度学习中又一个奇特现象，此外还有grokking和双重下降（double descent）。（2）分叉是自动研究代理提出的技巧所产生的意外且令人不快的副产品。虽然这些代理产生了大量看似有用的结果，但我们应始终对其结果保持谨慎。
英文摘要
This paper studies forking, a generalization failure discovered in NanoGPT autoresearch. Under data replay, models with an over-encoding n-gram memory branch show a sharp separation of training and validation loss at epoch boundaries, resembling the shape of forks. We study this phenomenon in a controlled vanilla NanoGPT setting and reproduce it in a DeepSeek-style model with Engram. Mechanistically, repeated updates sharpen the continuations observed in training while suppressing the probability of unseen continuations, whose loss grows with each pass. The n-gram module creates weakly interacting context-specific subspaces, amplifying this effect. Low-frequency contexts contribute most of the gap, whereas larger training budgets and heavily crowded tables suppress it. We also observe forking in short-budget, heavily repeated SFT and RL-like regimes. The contributions of this paper are twofold: (1) Forking reveals yet another curious phenomenon in deep learning, in addition to grokking and double descent. (2) Forking is an unexpected and unpleasant by-product of tricks proposed by autoresearch agents. While these agents produce an enormous number of results that seem useful, we should always be careful with their results.
Comments42 pages, 22 figures. Code and reproduction materials: https://github.com/guoshaoyang-pku/forking/tree/release
