SLLM-multi-hop/LanguageQA

Dataset Card for SAKURA-LanguageQA This dataset contains the audio and the single/multi-hop questions/answers of the language track of the SAKURA benchmark from Interspeech 2025 paper, "SAKURA: On the Multi-hop Reasoning of Large Audio-Language Models Based on Speech and Audio Information". The fields of the dataset are: file: The filename of the audio files. audio: The audio recordings. attribute_label: The attribute labels (i.e., the language spoken in the speech) of the… See the full description on the dataset page: https://huggingface.co/datasets/SLLM-multi-hop/LanguageQA.