Disentangling language understanding and reasoning structures in cross-lingual chain-of-thought prompting

Simple item page
dc.contributor.authorTran , Khanh-Tungen
dc.contributor.authorVu, Nguyet-Hangen
dc.contributor.authorO'Sullivan, Barryen
dc.contributor.authorNguyen, Hoang D.en
dc.contributor.funderResearch Irelanden
dc.date.accessioned2025-12-18T12:00:23Z
dc.date.available2025-12-18T12:00:23Z
dc.date.issued2025-11en
dc.description.abstractCross-lingual chain-of-thought prompting techniques have proven effective for investigating diverse reasoning paths in Large Language Models (LLMs), especially for low-resource languages. Despite these empirical gains, the mechanisms underlying cross-lingual improvements remain perplexing. This study, therefore, addresses whether the benefits of cross-lingual prompting arise from reasoning structures intrinsic to each language, or are simply a consequence of improved comprehension through cross-linguistic exposure. We employ neuron intervention and perturbation techniques to analyze and deactivate language-specific reasoning neurons during cross-lingual prompting, leading to performance disparities across languages, up to 27.4%. Our findings disentangle that these neurons are essential for reasoning in their respective languages but have minimal effect on reasoning in other languages, providing evidence for the existence of language-specific local reasoning structures and guiding the development of more interpretable and effective multilingual AI systems.en
dc.description.sponsorshipResearch Ireland (12/RC/2289-P2)en
dc.description.statusPeer revieweden
dc.description.versionAccepted Versionen
dc.format.mimetypeapplication/pdfen
dc.identifier.citationTran, K.-T., Vu, N.-H., O'Sullivan, B. and Nguyen, H. D. (2025) 'Disentangling language understanding and reasoning structures in cross-lingual chain-of-thought prompting', Findings of the Association for Computational Linguistics: EMNLP 2025, pp. 12200-12206. https://doi.org/10.18653/v1/2025.findings-emnlp.652en
dc.identifier.doi10.18653/v1/2025.findings-emnlp.652en
dc.identifier.endpage12206en
dc.identifier.isbn979-8-89176-335-7en
dc.identifier.journaltitleFindings of the Association for Computational Linguisticsen
dc.identifier.startpage12200en
dc.identifier.urihttps://hdl.handle.net/10468/18353
dc.identifier.volumeEMNLP 2025en
dc.language.isoenen
dc.publisherAssociation for Computational Linguisticsen
dc.relation.ispartofEMNLP 2025, Suzhou, China, 4-9 November 2025en
dc.relation.projectinfo:eu-repo/grantAgreement/SFI/Centres for Research Training (CRT) Programme/18/CRT/6223/IE/SFI Centre for Research Training in Artificial Intelligence/en
dc.relation.project12/RC/2289-P2en
dc.rights© 2025, Association for Computational Linguistics.en
dc.rights.urihttps://creativecommons.org/licenses/by/4.0/en
dc.subjectCross-lingual chain-of-thought prompting techniquesen
dc.subjectLarge Language Models (LLMs)en
dc.titleDisentangling language understanding and reasoning structures in cross-lingual chain-of-thought promptingen
dc.typeConference itemen
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
2025_EMNLP_CoT_Prompting.pdf
Size:
595.53 KB
Format:
Adobe Portable Document Format
Description:
Accepted Version
Download
License bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
license.txt
Size:
2.71 KB
Format:
Item-specific license agreed upon to submission
Description:
Download
Collections
Computer Science - Conference Items
ADVANCE CRT - Conference Items
Insight Centre for Data Analytics - Conference Items