@inproceedings{muszynski-etal-2026-mllm,
    title = "mllm-shap: A Shapley Value Explainability Platform for Text-Audio Multimodal Large Language Models",
    author = "Muszy{\'n}ski, Jakub  and
      Pozorski, Pawe{\l}  and
      Ganzha, Maria",
    editor = "Durrett, Greg  and
      Jian, Ping",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 3: System Demonstrations)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-demo.38/",
    pages = "387--396",
    ISBN = "979-8-89176-392-0",
    abstract = "We present mllm-shap, an open-sourcePython platform for researchers and ML practitioners that extends Shapley value (SV)explainability from text-only large languagemodels to multimodal LLMs (MLLMs) thatjointly process text and audio. Buildingon the token-level SV framework introducedby TokenSHAP, mllm-shap addresses threechallenges absent in the text-only setting:(1) modality-aware coalition masking thathandles the coexistence of text tokens anddense audio encoder frames within a single input, (2) multi-turn conversation tracking withper-token role and modality metadata, and(3) audio token grouping via phonetic alignment that reduces the coalition space by 10{--}50 times. The platform ships as a pip-installablepackage implementing five SV estimation strategies {--} including a Complementary Contributions estimator with Neyman-optimal allocation that outperforms Monte Carlo baselines {--} together with an interactive web GUIfor real-time attribution visualization. Toour knowledge, mllm-shap is the first publicly available framework for complete, reproducible SV-based explainability of text-audioMLLMs. The package is MIT-licensed withfull source code on GitHub and a demonstration video included as supplementary material."
}