How Can We Know <i>When</i> Language Models Know? On the Calibration of Language Models for Question Answering

TLDR

Language models capture factual and commonsense knowledge but still frequently provide incorrect answers. The study asks how to determine when a language model’s confidence truly reflects correctness, focusing on calibration for question answering. The authors evaluate calibration of T5, BART, and GPT‑2 on QA tasks, then apply fine‑tuning, post‑hoc probability adjustments, and output/input modifications to improve confidence‑correctness alignment, and analyze the strengths and limits of these techniques. Across diverse datasets the calibration methods markedly improve confidence‑correctness correlation, and the authors release code to support further research.

Abstract

Abstract Recent works have shown that language models (LM) capture different types of knowledge regarding facts or common sense. However, because no model is perfect, they still fail to provide appropriate answers in many cases. In this paper, we ask the question, “How can we know when language models know, with confidence, the answer to a particular query?” We examine this question from the point of view of calibration, the property of a probabilistic model’s predicted probabilities actually being well correlated with the probabilities of correctness. We examine three strong generative models—T5, BART, and GPT-2—and study whether their probabilities on QA tasks are well calibrated, finding the answer is a relatively emphatic no. We then examine methods to calibrate such models to make their confidence scores correlate better with the likelihood of correctness through fine-tuning, post-hoc probability modification, or adjustment of the predicted outputs or inputs. Experiments on a diverse range of datasets demonstrate the effectiveness of our methods. We also perform analysis to study the strengths and limitations of these methods, shedding light on further improvements that may be made in methods for calibrating LMs. We have released the code at https://github.com/jzbjyb/lm-calibration.