A Study on the Calibration of In-context Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Hanlin, Zhang, Yi-Fan, Yu, Yaodong, Madeka, Dhruv, Foster, Dean, Xing, Eric, Lakkaraju, Himabindu, Kakade, Sham
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910387739295744
author Zhang, Hanlin
Zhang, Yi-Fan
Yu, Yaodong
Madeka, Dhruv
Foster, Dean
Xing, Eric
Lakkaraju, Himabindu
Kakade, Sham
author_facet Zhang, Hanlin
Zhang, Yi-Fan
Yu, Yaodong
Madeka, Dhruv
Foster, Dean
Xing, Eric
Lakkaraju, Himabindu
Kakade, Sham
contents Accurate uncertainty quantification is crucial for the safe deployment of machine learning models, and prior research has demonstrated improvements in the calibration of modern language models (LMs). We study in-context learning (ICL), a prevalent method for adapting static LMs through tailored prompts, and examine the balance between performance and calibration across a broad spectrum of natural language understanding and reasoning tasks. Through comprehensive experiments, we observe that, with an increasing number of ICL examples, models initially exhibit increased miscalibration before achieving better calibration and miscalibration tends to arise in low-shot settings. Moreover, we find that methods aimed at improving usability, such as fine-tuning and chain-of-thought (CoT) prompting, can lead to miscalibration and unreliable natural language explanations. Furthermore, we explore recalibration techniques and find that a scaling-binning calibrator can reduce calibration errors consistently.
format Preprint
id arxiv_https___arxiv_org_abs_2312_04021
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle A Study on the Calibration of In-context Learning
Zhang, Hanlin
Zhang, Yi-Fan
Yu, Yaodong
Madeka, Dhruv
Foster, Dean
Xing, Eric
Lakkaraju, Himabindu
Kakade, Sham
Computation and Language
Artificial Intelligence
Machine Learning
Accurate uncertainty quantification is crucial for the safe deployment of machine learning models, and prior research has demonstrated improvements in the calibration of modern language models (LMs). We study in-context learning (ICL), a prevalent method for adapting static LMs through tailored prompts, and examine the balance between performance and calibration across a broad spectrum of natural language understanding and reasoning tasks. Through comprehensive experiments, we observe that, with an increasing number of ICL examples, models initially exhibit increased miscalibration before achieving better calibration and miscalibration tends to arise in low-shot settings. Moreover, we find that methods aimed at improving usability, such as fine-tuning and chain-of-thought (CoT) prompting, can lead to miscalibration and unreliable natural language explanations. Furthermore, we explore recalibration techniques and find that a scaling-binning calibrator can reduce calibration errors consistently.
title A Study on the Calibration of In-context Learning
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2312.04021