@inproceedings{zhang-etal-2026-calibrating,
title = "Calibrating Student Models via Uncertainty-Integrated Distillation",
author = "Zhang, Ruixuan and
Tan, Hongye and
Hu, Zhiwei and
Han, XiaoQi and
Li, Ru and
Pan, Jeff Z. and
Gutierrez Basulto, Victor",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1930/",
pages = "43923--43938",
ISBN = "979-8-89176-459-0",
abstract = "Model calibration aligns a model{'}s confidence with the accuracy of its predictions, directly influencing reliability and user trust. Surprisingly, existing research on knowledge distillation has paid little attention to the calibration of student models. In this paper, we evaluate calibration across diverse knowledge distillation settings and demonstrate that student models distilled from black-box teachers generally exhibit poor calibration. We attribute this to the fact that distillation fails to convey well-calibrated confidence, which serves as a crucial expression of predictive uncertainty. To address this limitation, we propose an ***U**ncertainty-**I**ntegrated **K**nowledge **D**istillation* (**UIKD**) method comprising two key components: (1) *uncertainty estimation*, which leverages the collective intelligence of multiple LLMs via a debate mechanism to quantify uncertainty in the distillation data; and (2) *uncertainty integration*, which employs an adaptive weighted loss function to integrate this data uncertainty into the student model effectively. Experiments on common-sense reasoning and natural language inference benchmarks demonstrate that UIKD consistently improves student model calibration while maintaining predictive accuracy. Our code is available at https://github.com/RuixuanZhang127/UIKD."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2026-calibrating">
<titleInfo>
<title>Calibrating Student Models via Uncertainty-Integrated Distillation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ruixuan</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hongye</namePart>
<namePart type="family">Tan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhiwei</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">XiaoQi</namePart>
<namePart type="family">Han</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ru</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jeff</namePart>
<namePart type="given">Z</namePart>
<namePart type="family">Pan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Victor</namePart>
<namePart type="family">Gutierrez Basulto</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Model calibration aligns a model’s confidence with the accuracy of its predictions, directly influencing reliability and user trust. Surprisingly, existing research on knowledge distillation has paid little attention to the calibration of student models. In this paper, we evaluate calibration across diverse knowledge distillation settings and demonstrate that student models distilled from black-box teachers generally exhibit poor calibration. We attribute this to the fact that distillation fails to convey well-calibrated confidence, which serves as a crucial expression of predictive uncertainty. To address this limitation, we propose an ***U**ncertainty-**I**ntegrated **K**nowledge **D**istillation* (**UIKD**) method comprising two key components: (1) *uncertainty estimation*, which leverages the collective intelligence of multiple LLMs via a debate mechanism to quantify uncertainty in the distillation data; and (2) *uncertainty integration*, which employs an adaptive weighted loss function to integrate this data uncertainty into the student model effectively. Experiments on common-sense reasoning and natural language inference benchmarks demonstrate that UIKD consistently improves student model calibration while maintaining predictive accuracy. Our code is available at https://github.com/RuixuanZhang127/UIKD.</abstract>
<identifier type="citekey">zhang-etal-2026-calibrating</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1930/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>43923</start>
<end>43938</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Calibrating Student Models via Uncertainty-Integrated Distillation
%A Zhang, Ruixuan
%A Tan, Hongye
%A Hu, Zhiwei
%A Han, XiaoQi
%A Li, Ru
%A Pan, Jeff Z.
%A Gutierrez Basulto, Victor
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F zhang-etal-2026-calibrating
%X Model calibration aligns a model’s confidence with the accuracy of its predictions, directly influencing reliability and user trust. Surprisingly, existing research on knowledge distillation has paid little attention to the calibration of student models. In this paper, we evaluate calibration across diverse knowledge distillation settings and demonstrate that student models distilled from black-box teachers generally exhibit poor calibration. We attribute this to the fact that distillation fails to convey well-calibrated confidence, which serves as a crucial expression of predictive uncertainty. To address this limitation, we propose an ***U**ncertainty-**I**ntegrated **K**nowledge **D**istillation* (**UIKD**) method comprising two key components: (1) *uncertainty estimation*, which leverages the collective intelligence of multiple LLMs via a debate mechanism to quantify uncertainty in the distillation data; and (2) *uncertainty integration*, which employs an adaptive weighted loss function to integrate this data uncertainty into the student model effectively. Experiments on common-sense reasoning and natural language inference benchmarks demonstrate that UIKD consistently improves student model calibration while maintaining predictive accuracy. Our code is available at https://github.com/RuixuanZhang127/UIKD.
%U https://aclanthology.org/2026.emnlp-main.1930/
%P 43923-43938
Markdown (Informal)
[Calibrating Student Models via Uncertainty-Integrated Distillation](https://aclanthology.org/2026.emnlp-main.1930/) (Zhang et al., EMNLP 2026)
ACL
- Ruixuan Zhang, Hongye Tan, Zhiwei Hu, XiaoQi Han, Ru Li, Jeff Z. Pan, and Victor Gutierrez Basulto. 2026. Calibrating Student Models via Uncertainty-Integrated Distillation. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 43923–43938, Budapest, Hungary. Association for Computational Linguistics.