Saved in:
Bibliographic Details
Main Authors: Zhou, Xiaoling, Ye, Wei, Wang, Yidong, Jiang, Chaoya, Lee, Zhemg, Xie, Rui, Zhang, Shikun
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2407.00100
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914852241408000
author Zhou, Xiaoling
Ye, Wei
Wang, Yidong
Jiang, Chaoya
Lee, Zhemg
Xie, Rui
Zhang, Shikun
author_facet Zhou, Xiaoling
Ye, Wei
Wang, Yidong
Jiang, Chaoya
Lee, Zhemg
Xie, Rui
Zhang, Shikun
contents The emergence of in-context learning (ICL) enables large pre-trained language models (PLMs) to make predictions for unseen inputs without updating parameters. Despite its potential, ICL's effectiveness heavily relies on the quality, quantity, and permutation of demonstrations, commonly leading to suboptimal and unstable performance. In this paper, we tackle this challenge for the first time from the perspective of demonstration augmentation. Specifically, we start with enriching representations of demonstrations by leveraging their deep feature distribution. We then theoretically reveal that when the number of augmented copies approaches infinity, the augmentation is approximately equal to a novel logit calibration mechanism integrated with specific statistical properties. This insight results in a simple yet highly efficient method that significantly improves the average and worst-case accuracy across diverse PLMs and tasks. Moreover, our method effectively reduces performance variance among varying demonstrations, permutations, and templates, and displays the capability to address imbalanced class distributions.
format Preprint
id arxiv_https___arxiv_org_abs_2407_00100
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Enhancing In-Context Learning via Implicit Demonstration Augmentation
Zhou, Xiaoling
Ye, Wei
Wang, Yidong
Jiang, Chaoya
Lee, Zhemg
Xie, Rui
Zhang, Shikun
Machine Learning
Artificial Intelligence
Computation and Language
I.2.7
The emergence of in-context learning (ICL) enables large pre-trained language models (PLMs) to make predictions for unseen inputs without updating parameters. Despite its potential, ICL's effectiveness heavily relies on the quality, quantity, and permutation of demonstrations, commonly leading to suboptimal and unstable performance. In this paper, we tackle this challenge for the first time from the perspective of demonstration augmentation. Specifically, we start with enriching representations of demonstrations by leveraging their deep feature distribution. We then theoretically reveal that when the number of augmented copies approaches infinity, the augmentation is approximately equal to a novel logit calibration mechanism integrated with specific statistical properties. This insight results in a simple yet highly efficient method that significantly improves the average and worst-case accuracy across diverse PLMs and tasks. Moreover, our method effectively reduces performance variance among varying demonstrations, permutations, and templates, and displays the capability to address imbalanced class distributions.
title Enhancing In-Context Learning via Implicit Demonstration Augmentation
topic Machine Learning
Artificial Intelligence
Computation and Language
I.2.7
url https://arxiv.org/abs/2407.00100