Skip to main navigation Skip to search Skip to main content

Tagged Span Annotation for Detecting Translation Errors in Reasoning LLMs

  • Sungkyunkwan University
  • Samsung

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

We present the submission of the AIP team to the WMT 2025 Unified MT Evaluation Shared Task, focusing on the span-level error detection subtask. Our system emphasizes response-format design to better harness the capabilities of OpenAI's o3, the state-of-the-art reasoning LLM. To this end, we introduce Tagged Span Annotation (TSA), an annotation scheme designed to more accurately extract span-level information from the LLM. On our refined version of WMT24 ESA dataset, our reference-free method achieves an F1 score of approximately 27 for character-level label prediction, outperforming the reference-based XCOMET-XXL at approximately 17.

Original languageEnglish
Title of host publicationWMT 2025 - 10th Conference on Machine Translation, Proceedings of the Conference
EditorsBarry Haddow, Tom Kocmi, Philipp Koehn, Christof Monz
PublisherAssociation for Computational Linguistics
Pages878-886
Number of pages9
ISBN (Electronic)9798891763418
DOIs
StatePublished - 2025
Event10th Conference on Machine Translation, WMT 2025 - Suzhou, China
Duration: 8 Nov 2025 → 9 Nov 2025

Publication series

NameConference on Machine Translation - Proceedings
ISSN (Electronic)2768-0983

Conference

Conference10th Conference on Machine Translation, WMT 2025
Country/TerritoryChina
CitySuzhou
Period8/11/25 → 9/11/25

Fingerprint

Dive into the research topics of 'Tagged Span Annotation for Detecting Translation Errors in Reasoning LLMs'. Together they form a unique fingerprint.

Cite this