@inproceedings{tian-etal-2025-distance,
title = "Distance between Relevant Information Pieces Causes Bias in Long-Context {LLM}s",
author = "Tian, Runchu and
Li, Yanghao and
Fu, Yuepeng and
Deng, Siyang and
Luo, Qinyu and
Qian, Cheng and
Wang, Shuo and
Cong, Xin and
Zhang, Zhong and
Wu, Yesai and
Lin, Yankai and
Wang, Huadong and
Liu, Xiaojiang",
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.28/",
doi = "10.18653/v1/2025.findings-acl.28",
pages = "521--533",
ISBN = "979-8-89176-256-5",
abstract = "Positional bias in large language models hinders their ability to effectively process long inputs. A prominent example is the ``lost in the middle'' phenomenon, where LLMs struggle to utilize relevant information situated in the middle of the input. While prior research primarily focuses on single pieces of relevant information, real-world applications often involve multiple relevant information pieces. To bridge this gap, we present LongPiBench, a benchmark designed to assess positional bias involving multiple pieces of relevant information. It includes various tasks and input lengths. Thorough experiments are conducted with three commercial and six open-source models. These experiments reveal that while most current models are more robust against the ``lost in the middle'' issue, there also exist noticeable biases related to the spacing of relevant information pieces. These findings highlight the importance of evaluating and reducing positional biases for long-context LLMs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="tian-etal-2025-distance">
<titleInfo>
<title>Distance between Relevant Information Pieces Causes Bias in Long-Context LLMs</title>
</titleInfo>
<name type="personal">
<namePart type="given">Runchu</namePart>
<namePart type="family">Tian</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yanghao</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuepeng</namePart>
<namePart type="family">Fu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Siyang</namePart>
<namePart type="family">Deng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qinyu</namePart>
<namePart type="family">Luo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Cheng</namePart>
<namePart type="family">Qian</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuo</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xin</namePart>
<namePart type="family">Cong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhong</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yesai</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yankai</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Huadong</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaojiang</namePart>
<namePart type="family">Liu</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>Positional bias in large language models hinders their ability to effectively process long inputs. A prominent example is the “lost in the middle” phenomenon, where LLMs struggle to utilize relevant information situated in the middle of the input. While prior research primarily focuses on single pieces of relevant information, real-world applications often involve multiple relevant information pieces. To bridge this gap, we present LongPiBench, a benchmark designed to assess positional bias involving multiple pieces of relevant information. It includes various tasks and input lengths. Thorough experiments are conducted with three commercial and six open-source models. These experiments reveal that while most current models are more robust against the “lost in the middle” issue, there also exist noticeable biases related to the spacing of relevant information pieces. These findings highlight the importance of evaluating and reducing positional biases for long-context LLMs.</abstract>
<identifier type="citekey">tian-etal-2025-distance</identifier>
<identifier type="doi">10.18653/v1/2025.findings-acl.28</identifier>
<location>
<url>https://aclanthology.org/2025.findings-acl.28/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>521</start>
<end>533</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Distance between Relevant Information Pieces Causes Bias in Long-Context LLMs
%A Tian, Runchu
%A Li, Yanghao
%A Fu, Yuepeng
%A Deng, Siyang
%A Luo, Qinyu
%A Qian, Cheng
%A Wang, Shuo
%A Cong, Xin
%A Zhang, Zhong
%A Wu, Yesai
%A Lin, Yankai
%A Wang, Huadong
%A Liu, Xiaojiang
%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 tian-etal-2025-distance
%X Positional bias in large language models hinders their ability to effectively process long inputs. A prominent example is the “lost in the middle” phenomenon, where LLMs struggle to utilize relevant information situated in the middle of the input. While prior research primarily focuses on single pieces of relevant information, real-world applications often involve multiple relevant information pieces. To bridge this gap, we present LongPiBench, a benchmark designed to assess positional bias involving multiple pieces of relevant information. It includes various tasks and input lengths. Thorough experiments are conducted with three commercial and six open-source models. These experiments reveal that while most current models are more robust against the “lost in the middle” issue, there also exist noticeable biases related to the spacing of relevant information pieces. These findings highlight the importance of evaluating and reducing positional biases for long-context LLMs.
%R 10.18653/v1/2025.findings-acl.28
%U https://aclanthology.org/2025.findings-acl.28/
%U https://doi.org/10.18653/v1/2025.findings-acl.28
%P 521-533
Markdown (Informal)
[Distance between Relevant Information Pieces Causes Bias in Long-Context LLMs](https://aclanthology.org/2025.findings-acl.28/) (Tian et al., Findings 2025)
ACL
- Runchu Tian, Yanghao Li, Yuepeng Fu, Siyang Deng, Qinyu Luo, Cheng Qian, Shuo Wang, Xin Cong, Zhong Zhang, Yesai Wu, Yankai Lin, Huadong Wang, and Xiaojiang Liu. 2025. Distance between Relevant Information Pieces Causes Bias in Long-Context LLMs. In Findings of the Association for Computational Linguistics: ACL 2025, pages 521–533, Vienna, Austria. Association for Computational Linguistics.