Label Words are Anchors: An Information Flow Perspective for Understanding In-Context Learning

TL;DR

This paper investigates how large language models perform in-context learning by viewing label words as anchors that facilitate information aggregation and prediction, leading to improved methods and diagnostic tools.

Contribution

It reveals the role of label words as anchors in ICL, introduces an anchor re-weighting method, a demonstration compression technique, and an analysis framework for diagnosing errors.

Findings

Label words act as semantic anchors during ICL.

Anchor re-weighting improves ICL performance.

Demonstration compression speeds up inference.

Abstract

In-context learning (ICL) emerges as a promising capability of large language models (LLMs) by providing them with demonstration examples to perform diverse tasks. However, the underlying mechanism of how LLMs learn from the provided context remains under-explored. In this paper, we investigate the working mechanism of ICL through an information flow lens. Our findings reveal that label words in the demonstration examples function as anchors: (1) semantic information aggregates into label word representations during the shallow computation layers' processing; (2) the consolidated information in label words serves as a reference for LLMs' final predictions. Based on these insights, we introduce an anchor re-weighting method to improve ICL performance, a demonstration compression technique to expedite inference, and an analysis framework for diagnosing ICL errors in GPT2-XL. The promising applications of our findings again validate the uncovered ICL working mechanism and pave the way for future studies.