SLLM-multi-hop/LanguageQA

SAKURA-LanguageQA数据集包含了基于语音和音频信息的SAKURA benchmark语言轨道中的音频文件及其单跳和多跳问题的问答数据。数据集字段包括音频文件名、音频数据、属性标签（即语音中的语言）、单跳问题指令、单跳答案、多跳问题指令和多跳答案。

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 include the filename of the audio files, the audio recordings, attribute labels (i.e., the language spoken in the speech) of the audio files, single-hop questions (instructions), single-hop answers, multi-hop questions (instructions), and multi-hop answers.