CORDIAL: Can Multimodal Large Language Models Effectively Understand Coherence Relationships?

CORDIAL is a new benchmark focused on Multimodal Discourse Analysis using Coherence Relations.

• Website
• Paper
• Dataset

News 🚀

• [2025-07-27] The benchmark dataset and code has been published!
• [2025-05-17] CORDIAL has been accepted to ACL (Main) 2025.
• [2025-02-16] Our paper is available on arxiv.

Citing

If you find our work useful, please consider citing:

@inproceedings{anantha-ramakrishnan-etal-2025-cordial,
    title = "{CORDIAL}: Can Multimodal Large Language Models Effectively Understand Coherence Relationships?",
    author = "Anantha Ramakrishnan, Aashish  and
      Ramakrishnan, Aadarsh Anantha  and
      Lee, Dongwon",
    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.1033/",
    doi = "10.18653/v1/2025.acl-long.1033",
    pages = "21277--21297",
    ISBN = "979-8-89176-251-0"
}