归纳推理任务中Transformer的不变学习动态

中文摘要 AI 辅助

我们提出一个理论框架来解释Transformer语言模型中归纳推理能力的出现。以往关于Transformer学习动态的工作大多局限于特定任务，我们研究了一类广义归纳任务，它统一了文献中已知的几个合成任务，包括上下文n元语法和多跳推理。在此类任务中，我们从理论上证明了注意力模型的训练动态可以被限制在一个高度可解释的低维不变流形上。在这个流形上，学习动态由少数可解释坐标而非数百万参数捕捉，使理论和实证分析更易处理。利用此框架，我们刻画了数据统计如何控制上下文学习和权重学习之间的竞争，研究了随机初始化如何在多个解决方案可行时确定“获胜”电路，还证明了与流形相关的坐标框架可用于自动检测训练模型中学习到的电路。通过将电路形成视为低维动态现象，我们朝着Transformer学习的预测理论迈进了一步。

英文摘要

We present a theoretical framework to explain the emergence of inductive reasoning abilities in Transformer language models. While previous works on Transformer learning dynamics have so far been mostly tied to specific tasks, we study a generalized class of inductive tasks that unifies several synthetic tasks known in the literature, including in-context n-grams and multi-hop reasoning. In this class, we theoretically prove that the training dynamics of attention models can be confined to a highly interpretable, low-dimensional invariant manifold. On this manifold, the learning dynamics are captured by a handful of interpretable coordinates rather than millions of parameters, making both theoretical and empirical analysis more tractable. Using this framework, we characterize how data statistics govern the competition between in-context and in-weights learning, we study how random initializations determine the `winning' circuit when multiple solutions are possible, and we demonstrate that the coordinate frame associated with the manifold can be used to automatically detect which circuits have been learned in trained models. By casting circuit formation as a low-dimensional dynamical phenomenon, we take a step toward a predictive theory of how Transformers learn.

发表机构

• ETH Zurich（苏黎世联邦理工学院）
• Stanford（斯坦福大学）

机构由 AI 辅助整理，请以论文原文为准。