In-Context Convergence of Transformers

Published: 07 Nov 2023, Last Modified: 13 Dec 2023M3L 2023 OralEveryoneRevisionsBibTeX
Keywords: Theoretical in-context learning, Softmax attention, Training dynamics, Convergence
TL;DR: We establish the convergence guarantee for transformers with softmax attention trained via gradient descent for in-context learning.
Abstract: Transformers have recently revolutionized many domains in modern machine learning and one salient discovery is their remarkable in-context learning capability, where models can solve an unseen task by utilizing task-specific prompts without further parameters fine-tuning. This also inspired recent theoretical studies aiming to understand the in-context learning mechanism of transformers, which however focused only on $\textbf{linear}$ transformers. In this work, we take the first step toward studying the learning dynamics of a one-layer transformer with $\textbf{softmax}$ attention trained via gradient descent in order to in-context learn linear function classes. We consider a structured data model, where each token is randomly sampled from a set of feature vectors in either balanced or imbalanced fashion. For data with balanced features, we establish the finite-time convergence guarantee with near-zero prediction error by navigating our analysis over two phases of the training dynamics of the attention map. More notably, for data with imbalanced features, we show that the learning dynamics take a stage-wise convergence process, where the transformer first converges to a near-zero prediction error for the query tokens of dominant features, and then converges later to a near-zero prediction error for the query tokens of under-represented features, respectively via one and four training phases. Our proof features new techniques for analyzing the competing strengths of two types of attention weights, the change of which determines different phases.
Submission Number: 27