Token-based Decision Criteria Are Suboptimal in In-context Learning

TL;DR

This paper introduces Hidden Calibration, a new classification method for in-context learning that replaces token probability-based criteria with a nearest centroid classifier on hidden states, significantly improving performance.

Contribution

It proposes Hidden Calibration, a novel approach that uses hidden state clustering instead of token probabilities, achieving state-of-the-art results in ICL tasks.

Findings

Hidden Calibration outperforms token-based methods by 20-50%.

LM hidden states form linearly separable intra-class clusters.

Better classification criteria reduce inter-class overlap.

Abstract

In-Context Learning (ICL) typically utilizes classification criteria from output probabilities of manually selected label tokens. However, we argue that such token-based classification criteria lead to suboptimal decision boundaries, despite delicate calibrations through translation and constrained rotation applied. To address this problem, we propose Hidden Calibration, which renounces token probabilities and uses the nearest centroid classifier on the LM's last hidden states. In detail, we assign the label of the nearest centroid previously estimated from a calibration set to the test sample as the predicted label. Our experiments on 6 models and 10 classification datasets indicate that Hidden Calibration consistently outperforms current token-based baselines by about 20%~50%, achieving a strong state-of-the-art in ICL. Our further analysis demonstrates that Hidden Calibration finds better classification criteria with less inter-class overlap, and LMs provide linearly separable intra-class clusters with the help of demonstrations, which supports Hidden Calibration and gives new insights into the principle of ICL. Our official code implementation can be found at https://github.com/hc495/Hidden_Calibration.