Is this your dataset? Claim it with the Hugging Face account that owns it.

ddwang2000/
MMSU

Name: MMSU
Creator: ddwang2000
License: https://spdx.org/licenses/MIT.html

Views

No views yet

	[ICLR 2026] MMSU: A Massive Multi-task Spoken Language Understanding and Reasoning Benchmark






	


	Overview of MMSU

MMSU (Massive Multi-task Spoken Language Understanding and Reasoning Benchmark) is a comprehensive benchmark for evaluating fine-grained spoken language understanding and reasoning in multimodal models. It systematically captures the variance of real-world linguistic phenomena in daily speech through 47 sub-tasks, including phonetics, prosody, rhetoric… See the full description on the dataset page: https://huggingface.co/datasets/ddwang2000/MMSU.