CoChat

The DOI matches its Crossref record; 72 of the 72 references it cites were found in a registry or an index.

How Can We Know When Language Models Know? On the Calibration of Language Models for Question Answering

Open access
Jiang, Zhengbao, Araki, Jun, Ding, Haibo, Neubig, Graham · 2021 · Transactions of the Association for Computational Linguistics

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.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1162/tacl_a_00407

Cites

72 of 72 references verified0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

8 more claims about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Research works printed but not linked: 4

  1. Hewitt, 2019, A structural probe for finding syntax in word representations, 4129
  2. Ng, 2019, Facebook fair’s WMT19 news translation task submission, 314
  3. Quach, 2020, Researchers made an OpenAI GPT-3 medical chatbot as an experiment. It told a mock patient to kill themselves, The Register
  4. Radford, 2019, Language models are unsupervised multitask learners, OpenAI Blog, 1

Cited by

See all 305

Metadata

Authors
Jiang, Z., Araki, J., Ding, H., & Neubig, G.
Affiliations
  • Carnegie Mellon University (US)
  • Robert Bosch (United States) (US)
Corresponding author
  • Jiang, Zhengbao
  • Araki, Jun
  • Ding, Haibo
  • Neubig, Graham
Venue
Transactions of the Association for Computational Linguistics
Publisher
MIT Press
Published
2021
Volume
9
Pages
962-977
ISSN
2307-387X (ISSN-L)
OpenAlex
W3199958362
Cited by
177
References
71
Citations by year
2026: 43 · 2025: 49 · 2024: 32 · 2023: 37 · 2022: 14 · 2021: 2
FWCI
7.48
Type
journal article
Language
en
Open access
Locations
  • Transactions of the Association for Computational Linguistics (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
  • Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
  • Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
Keywords
question answering, confidence calibration, probabilistic reasoning, T5, BART, GPT-2, fine-tuning, knowledge capture, factual knowledge, commonsense knowledge

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.