@inproceedings{kapoor-etal-2024-calibration,
title = "Calibration-Tuning: Teaching Large Language Models to Know What They Don`t Know",
author = "Kapoor, Sanyam and
Gruver, Nate and
Roberts, Manley and
Pal, Arka and
Dooley, Samuel and
Goldblum, Micah and
Wilson, Andrew",
editor = {V{\'a}zquez, Ra{\'u}l and
Celikkanat, Hande and
Ulmer, Dennis and
Tiedemann, J{\"o}rg and
Swayamdipta, Swabha and
Aziz, Wilker and
Plank, Barbara and
Baan, Joris and
de Marneffe, Marie-Catherine},
booktitle = "Proceedings of the 1st Workshop on Uncertainty-Aware NLP (UncertaiNLP 2024)",
month = mar,
year = "2024",
address = "St Julians, Malta",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.uncertainlp-1.1/",
pages = "1--14",
abstract = "Large language models are increasingly deployed for high-stakes decision making, for example in financial and medical applications. In such applications, it is imperative that we be able to estimate our confidence in the answers output by a language model in order to assess risks. Although we can easily compute the probability assigned by a language model to the sequence of tokens that make up an answer, we cannot easily compute the probability of the answer itself, which could be phrased in numerous ways.While other works have engineered ways of assigning such probabilities to LLM outputs, a key problem remains: existing language models are poorly calibrated, often confident when they are wrong or unsure when they are correct. In this work, we devise a protocol called *calibration tuning* for finetuning LLMs to output calibrated probabilities. Calibration-tuned models demonstrate superior calibration performance compared to existing language models on a variety of question-answering tasks, including open-ended generation, without affecting accuracy. We further show that this ability transfers to new domains outside of the calibration-tuning train set."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kapoor-etal-2024-calibration">
<titleInfo>
<title>Calibration-Tuning: Teaching Large Language Models to Know What They Don‘t Know</title>
</titleInfo>
<name type="personal">
<namePart type="given">Sanyam</namePart>
<namePart type="family">Kapoor</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nate</namePart>
<namePart type="family">Gruver</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Manley</namePart>
<namePart type="family">Roberts</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Arka</namePart>
<namePart type="family">Pal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Samuel</namePart>
<namePart type="family">Dooley</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Micah</namePart>
<namePart type="family">Goldblum</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andrew</namePart>
<namePart type="family">Wilson</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-03</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 1st Workshop on Uncertainty-Aware NLP (UncertaiNLP 2024)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Raúl</namePart>
<namePart type="family">Vázquez</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hande</namePart>
<namePart type="family">Celikkanat</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dennis</namePart>
<namePart type="family">Ulmer</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jörg</namePart>
<namePart type="family">Tiedemann</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Swabha</namePart>
<namePart type="family">Swayamdipta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wilker</namePart>
<namePart type="family">Aziz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Barbara</namePart>
<namePart type="family">Plank</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joris</namePart>
<namePart type="family">Baan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marie-Catherine</namePart>
<namePart type="family">de Marneffe</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">St Julians, Malta</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Large language models are increasingly deployed for high-stakes decision making, for example in financial and medical applications. In such applications, it is imperative that we be able to estimate our confidence in the answers output by a language model in order to assess risks. Although we can easily compute the probability assigned by a language model to the sequence of tokens that make up an answer, we cannot easily compute the probability of the answer itself, which could be phrased in numerous ways.While other works have engineered ways of assigning such probabilities to LLM outputs, a key problem remains: existing language models are poorly calibrated, often confident when they are wrong or unsure when they are correct. In this work, we devise a protocol called *calibration tuning* for finetuning LLMs to output calibrated probabilities. Calibration-tuned models demonstrate superior calibration performance compared to existing language models on a variety of question-answering tasks, including open-ended generation, without affecting accuracy. We further show that this ability transfers to new domains outside of the calibration-tuning train set.</abstract>
<identifier type="citekey">kapoor-etal-2024-calibration</identifier>
<location>
<url>https://aclanthology.org/2024.uncertainlp-1.1/</url>
</location>
<part>
<date>2024-03</date>
<extent unit="page">
<start>1</start>
<end>14</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Calibration-Tuning: Teaching Large Language Models to Know What They Don‘t Know
%A Kapoor, Sanyam
%A Gruver, Nate
%A Roberts, Manley
%A Pal, Arka
%A Dooley, Samuel
%A Goldblum, Micah
%A Wilson, Andrew
%Y Vázquez, Raúl
%Y Celikkanat, Hande
%Y Ulmer, Dennis
%Y Tiedemann, Jörg
%Y Swayamdipta, Swabha
%Y Aziz, Wilker
%Y Plank, Barbara
%Y Baan, Joris
%Y de Marneffe, Marie-Catherine
%S Proceedings of the 1st Workshop on Uncertainty-Aware NLP (UncertaiNLP 2024)
%D 2024
%8 March
%I Association for Computational Linguistics
%C St Julians, Malta
%F kapoor-etal-2024-calibration
%X Large language models are increasingly deployed for high-stakes decision making, for example in financial and medical applications. In such applications, it is imperative that we be able to estimate our confidence in the answers output by a language model in order to assess risks. Although we can easily compute the probability assigned by a language model to the sequence of tokens that make up an answer, we cannot easily compute the probability of the answer itself, which could be phrased in numerous ways.While other works have engineered ways of assigning such probabilities to LLM outputs, a key problem remains: existing language models are poorly calibrated, often confident when they are wrong or unsure when they are correct. In this work, we devise a protocol called *calibration tuning* for finetuning LLMs to output calibrated probabilities. Calibration-tuned models demonstrate superior calibration performance compared to existing language models on a variety of question-answering tasks, including open-ended generation, without affecting accuracy. We further show that this ability transfers to new domains outside of the calibration-tuning train set.
%U https://aclanthology.org/2024.uncertainlp-1.1/
%P 1-14
Markdown (Informal)
[Calibration-Tuning: Teaching Large Language Models to Know What They Don’t Know](https://aclanthology.org/2024.uncertainlp-1.1/) (Kapoor et al., UncertaiNLP 2024)
ACL