Findings of the Fifth Shared Task on Multilingual Coreference Resolution: Expanding Datasets for Long-Range Entities

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

The Fifth Shared Task on Multilingual Coreference Resolution, conducted at the CODI-CRAC 2026 workshop, focused on advancing mention identification and identity-based coreference clustering. This 2026 edition specifically emphasized long-range entities, defined as coreferential chains extending across many words and sentences. The task significantly expanded its linguistic coverage by integrating five new datasets and two additional languages, all leveraging version 1.4 of CorefUD, a harmonized multilingual collection encompassing 27 datasets in 19 languages. Ten systems participated in total, including four LLM-based approaches, comprising three fine-tuned models and one few-shot method. While traditional systems maintained their performance lead, the LLM-based solutions demonstrated substantial potential, indicating they are poised to challenge established methods in future iterations of the task.

Key takeaway

For NLP Engineers developing multilingual coreference resolution systems, the CODI-CRAC 2026 shared task findings highlight the growing importance of long-range entity handling. You should prioritize integrating techniques capable of resolving coreference across significant textual distances. Consider experimenting with fine-tuned LLM-based approaches; they show promising potential to soon surpass traditional methods, offering a competitive edge in future system designs.

Key insights

Multilingual coreference resolution is advancing, with LLMs showing strong potential for long-range entity identification.

Principles

• Coreference resolution benefits from expanded multilingual datasets.
• Long-range coreference is a key challenge in NLP.
• LLMs are emerging as competitive in coreference tasks.

Method

The shared task required systems for mention identification and identity-based coreference clustering, evaluated on expanded multilingual datasets, including long-range entities.

In practice

• Explore CorefUD version 1.4 for multilingual coreference data.
• Investigate fine-tuned LLMs for long-range coreference.

Topics

• Multilingual Coreference Resolution
• Long-Range Entities
• Large Language Models
• CorefUD Dataset
• Natural Language Processing
• Shared Task Evaluation

Best for: Research Scientist, AI Scientist, NLP Engineer, Machine Learning Engineer

Related on AIssential

• A Larger Annotated Corpus of Portuguese Coreference — LLMs can annotate Portuguese coreference, but a new LLM-based system did not outperform existing rule-based methods.
• CorPipe at CRAC 2026: Empty Nodes and Cross-Lingual Transfer in Multilingual Coreference Resolution — CorPipe 26 integrates empty node prediction with mention and coreference resolution for improved multilingual performance.
• IndicDB -- Benchmarking Multilingual Text-to-SQL Capabilities in Indian Languages — IndicDB reveals a significant "Indic Gap" in LLM Text-to-SQL performance for Indian languages compared to English.
• Event and Entity Coreference Across Five Languages: Effects of Context and Referring Expression — Crosslingual coreference biases show personal pronouns favor entities, demonstratives favor events, influenced by antecedent complexity.
• Cross-Lingual Empirical Evaluation of Large Language Models for Arabic Medical Tasks — English-centric LLMs exhibit significant performance gaps and reliability issues in Arabic medical tasks.
• Coreference Resolution in spaCy — spaCy's experimental "coref" component enables robust coreference resolution for NLP applications.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.