@inproceedings{wang-etal-2025-tracing,
title = "Tracing and Dissecting How {LLM}s Recall Factual Knowledge for Real World Questions",
author = "Wang, Yiqun and
Wan, Chaoqun and
Hu, Sile and
Zhang, Yonggang and
Tian, Xiang and
Chen, Yaowu and
Shen, Xu and
Ye, Jieping",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.acl-long.1133/",
doi = "10.18653/v1/2025.acl-long.1133",
pages = "23246--23271",
ISBN = "979-8-89176-251-0",
abstract = "Recent advancements in large language models (LLMs) have shown promising ability to perform commonsense reasoning, bringing machines closer to human-like understanding. However, deciphering the internal reasoning processes of LLMs remains challenging due to the complex interdependencies among generated tokens, especially in practical question-answering. In this study, we introduce a two-dimensional analysis framework{---}comprising token back-tracing and individual token decoding{---}to uncover how LLMs conduct factual knowledge recall. Through explanatory analysis of three typical reasoning datasets, we identify a consistent three-phase pattern: Subject Augmentation and Broadcasting, Object Retrieval and Reranking, and Conclusion Fusion and Generation. Our findings reveal that LLMs do not lack relevant knowledge but struggle to select the most accurate information based on context during the retrieval and rerank phase. Leveraging these findings, we apply representation engineering and selective fine-tuning to target specific modules responsible for retrieval and rerank errors. Experimental results show large improvements in response accuracy for both in-domain and out-of-domain settings, validating the rationality of the interpreting result."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2025-tracing">
<titleInfo>
<title>Tracing and Dissecting How LLMs Recall Factual Knowledge for Real World Questions</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yiqun</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chaoqun</namePart>
<namePart type="family">Wan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sile</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yonggang</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiang</namePart>
<namePart type="family">Tian</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yaowu</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xu</namePart>
<namePart type="family">Shen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jieping</namePart>
<namePart type="family">Ye</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>Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</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">979-8-89176-251-0</identifier>
</relatedItem>
<abstract>Recent advancements in large language models (LLMs) have shown promising ability to perform commonsense reasoning, bringing machines closer to human-like understanding. However, deciphering the internal reasoning processes of LLMs remains challenging due to the complex interdependencies among generated tokens, especially in practical question-answering. In this study, we introduce a two-dimensional analysis framework—comprising token back-tracing and individual token decoding—to uncover how LLMs conduct factual knowledge recall. Through explanatory analysis of three typical reasoning datasets, we identify a consistent three-phase pattern: Subject Augmentation and Broadcasting, Object Retrieval and Reranking, and Conclusion Fusion and Generation. Our findings reveal that LLMs do not lack relevant knowledge but struggle to select the most accurate information based on context during the retrieval and rerank phase. Leveraging these findings, we apply representation engineering and selective fine-tuning to target specific modules responsible for retrieval and rerank errors. Experimental results show large improvements in response accuracy for both in-domain and out-of-domain settings, validating the rationality of the interpreting result.</abstract>
<identifier type="citekey">wang-etal-2025-tracing</identifier>
<identifier type="doi">10.18653/v1/2025.acl-long.1133</identifier>
<location>
<url>https://aclanthology.org/2025.acl-long.1133/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>23246</start>
<end>23271</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Tracing and Dissecting How LLMs Recall Factual Knowledge for Real World Questions
%A Wang, Yiqun
%A Wan, Chaoqun
%A Hu, Sile
%A Zhang, Yonggang
%A Tian, Xiang
%A Chen, Yaowu
%A Shen, Xu
%A Ye, Jieping
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-251-0
%F wang-etal-2025-tracing
%X Recent advancements in large language models (LLMs) have shown promising ability to perform commonsense reasoning, bringing machines closer to human-like understanding. However, deciphering the internal reasoning processes of LLMs remains challenging due to the complex interdependencies among generated tokens, especially in practical question-answering. In this study, we introduce a two-dimensional analysis framework—comprising token back-tracing and individual token decoding—to uncover how LLMs conduct factual knowledge recall. Through explanatory analysis of three typical reasoning datasets, we identify a consistent three-phase pattern: Subject Augmentation and Broadcasting, Object Retrieval and Reranking, and Conclusion Fusion and Generation. Our findings reveal that LLMs do not lack relevant knowledge but struggle to select the most accurate information based on context during the retrieval and rerank phase. Leveraging these findings, we apply representation engineering and selective fine-tuning to target specific modules responsible for retrieval and rerank errors. Experimental results show large improvements in response accuracy for both in-domain and out-of-domain settings, validating the rationality of the interpreting result.
%R 10.18653/v1/2025.acl-long.1133
%U https://aclanthology.org/2025.acl-long.1133/
%U https://doi.org/10.18653/v1/2025.acl-long.1133
%P 23246-23271
Markdown (Informal)
[Tracing and Dissecting How LLMs Recall Factual Knowledge for Real World Questions](https://aclanthology.org/2025.acl-long.1133/) (Wang et al., ACL 2025)
ACL
- Yiqun Wang, Chaoqun Wan, Sile Hu, Yonggang Zhang, Xiang Tian, Yaowu Chen, Xu Shen, and Jieping Ye. 2025. Tracing and Dissecting How LLMs Recall Factual Knowledge for Real World Questions. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 23246–23271, Vienna, Austria. Association for Computational Linguistics.