ππ MTR-DuplexBench has been accepted by ACL 2026 Findings!
π Paper | π€ HuggingFace Dataset
Affiliations: Tsinghua University, The Chinese University of Hong Kong, Huawei
This dataset is constructed to evaluate multi-modal audio models across four critical dimensions: Conversational Features, Instruction Following, Safety⦠See the full description on the dataset page:
https://huggingface.co/datasets/Jeff0918/MTR-DuplexBench.