@inproceedings{li-etal-2025-know,
title = "Know the Unknown: An Uncertainty-Sensitive Method for {LLM} Instruction Tuning",
author = "Li, Jiaqi and
Tang, Yixuan and
Yang, Yi",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-acl.153/",
doi = "10.18653/v1/2025.findings-acl.153",
pages = "2972--2989",
ISBN = "979-8-89176-256-5",
abstract = "Large language models (LLMs) demonstrate remarkable capabilities but face challenges from hallucinations, which typically arise from insufficient knowledge or context. While instructing LLMs to acknowledge knowledge limitations by responding with ``I don{'}t know'' appears promising, we find that models consistently struggle with admitting knowledge gaps. This challenge may originate from current instruction datasets that emphasise answer generation over knowledge boundary awareness. To address this limitation, we introduce **U**ncertainty-and-**S**ensitivity-Aware Tuning **(US-Tuning)**, a novel two-stage approach for contextual question answering (QA). The first stage enhances LLMs' ability to recognise their knowledge boundaries, while the second stage reinforces instruction adherence through carefully designed causal prompts. Our experimental results demonstrate that US-Tuning not only significantly reduces incorrect answers in contextual QA but also improves models' faithfulness to their parametric knowledge, mitigating hallucinations in general QA tasks. Our fine-tuned Llama2-7B model achieves up to a 34.7{\%} improvement in handling out-of-knowledge questions and outperforms GPT-4 by 4.2{\%} in overall performance."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="li-etal-2025-know">
<titleInfo>
<title>Know the Unknown: An Uncertainty-Sensitive Method for LLM Instruction Tuning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jiaqi</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yixuan</namePart>
<namePart type="family">Tang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yi</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-256-5</identifier>
</relatedItem>
<abstract>Large language models (LLMs) demonstrate remarkable capabilities but face challenges from hallucinations, which typically arise from insufficient knowledge or context. While instructing LLMs to acknowledge knowledge limitations by responding with “I don’t know” appears promising, we find that models consistently struggle with admitting knowledge gaps. This challenge may originate from current instruction datasets that emphasise answer generation over knowledge boundary awareness. To address this limitation, we introduce **U**ncertainty-and-**S**ensitivity-Aware Tuning **(US-Tuning)**, a novel two-stage approach for contextual question answering (QA). The first stage enhances LLMs’ ability to recognise their knowledge boundaries, while the second stage reinforces instruction adherence through carefully designed causal prompts. Our experimental results demonstrate that US-Tuning not only significantly reduces incorrect answers in contextual QA but also improves models’ faithfulness to their parametric knowledge, mitigating hallucinations in general QA tasks. Our fine-tuned Llama2-7B model achieves up to a 34.7% improvement in handling out-of-knowledge questions and outperforms GPT-4 by 4.2% in overall performance.</abstract>
<identifier type="citekey">li-etal-2025-know</identifier>
<identifier type="doi">10.18653/v1/2025.findings-acl.153</identifier>
<location>
<url>https://aclanthology.org/2025.findings-acl.153/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>2972</start>
<end>2989</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Know the Unknown: An Uncertainty-Sensitive Method for LLM Instruction Tuning
%A Li, Jiaqi
%A Tang, Yixuan
%A Yang, Yi
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Findings of the Association for Computational Linguistics: ACL 2025
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-256-5
%F li-etal-2025-know
%X Large language models (LLMs) demonstrate remarkable capabilities but face challenges from hallucinations, which typically arise from insufficient knowledge or context. While instructing LLMs to acknowledge knowledge limitations by responding with “I don’t know” appears promising, we find that models consistently struggle with admitting knowledge gaps. This challenge may originate from current instruction datasets that emphasise answer generation over knowledge boundary awareness. To address this limitation, we introduce **U**ncertainty-and-**S**ensitivity-Aware Tuning **(US-Tuning)**, a novel two-stage approach for contextual question answering (QA). The first stage enhances LLMs’ ability to recognise their knowledge boundaries, while the second stage reinforces instruction adherence through carefully designed causal prompts. Our experimental results demonstrate that US-Tuning not only significantly reduces incorrect answers in contextual QA but also improves models’ faithfulness to their parametric knowledge, mitigating hallucinations in general QA tasks. Our fine-tuned Llama2-7B model achieves up to a 34.7% improvement in handling out-of-knowledge questions and outperforms GPT-4 by 4.2% in overall performance.
%R 10.18653/v1/2025.findings-acl.153
%U https://aclanthology.org/2025.findings-acl.153/
%U https://doi.org/10.18653/v1/2025.findings-acl.153
%P 2972-2989
Markdown (Informal)
[Know the Unknown: An Uncertainty-Sensitive Method for LLM Instruction Tuning](https://aclanthology.org/2025.findings-acl.153/) (Li et al., Findings 2025)
ACL