AIR-Bench

# AIR-Bench Arxiv: https://arxiv.org/html/2402.07729v1 This is the AIR-Bench dataset download page. AIR-Bench encompasses two dimensions: foundation and chat benchmarks. The former consists of 19 tasks with approximately 19k single-choice questions. The latter one contains 2k instances of open-ended question-and-answer data. For how to run AIR-Bench, Please refer to AIR-Bench github page(https://github.com/OFA-Sys/AIR-Bench)(will be public soon). # Data Sources(All come from the Val/Test Set) - Musiccaps(https://huggingface.co/datasets/google/MusicCaps) under its license(https://choosealicense.com/licenses/cc-by-sa-4.0/). Copyright belongs to the original dataset owner. - Clotho(https://zenodo.org/records/3490684) under its license(Tampere University licence). Copyright belongs to the original dataset owner. - Fisher(https://catalog.ldc.upenn.edu/LDC2004T19) under its license(LDC User Agreement for Non-Members. Copyright belongs to the original dataset owner. - SpokenWOZ(https://spokenwoz.github.io/) under its license(https://creativecommons.org/licenses/by-nc/4.0/legalcode). Copyright belongs to the original dataset owner. - common_voice_13_0(https://huggingface.co/datasets/mozilla-foundation/common_voice_13_0) under its license(https://choosealicense.com/licenses/cc0-1.0/). Copyright belongs to the original dataset owner. - IEMOCAP(https://sail.usc.edu/iemocap/) under its license(https://sail.usc.edu/iemocap/Data_Release_Form_IEMOCAP.pdf). Copyright belongs to the original dataset owner. - CochlScene(https://zenodo.org/records/7080122) under its license(https://creativecommons.org/licenses/by-sa/3.0/legalcode). Copyright belongs to the original dataset owner. - TUT Acoustic scenes 2017(https://zenodo.org/records/400515) under its license(Other (Non-Commercial)). Copyright belongs to the original dataset owner. - Audio Grounding(https://github.com/wsntxxn/TextToAudioGrounding) under its license(https://github.com/wsntxxn/TextToAudioGrounding?tab=MIT-1-ov-file). Copyright belongs to the original dataset owner. - MUSIC-AVQA(https://gewu-lab.github.io/MUSIC-AVQA/) under its license(https://creativecommons.org/licenses/by-nc/4.0/). Copyright belongs to the original dataset owner. - FMA(https://github.com/mdeff/fma) under its license(https://creativecommons.org/licenses/by/4.0). Copyright belongs to the original dataset owner. - MTG-Jamendo(https://mtg.github.io/mtg-jamendo-dataset/) under its license(Creative Commons licenses). Copyright belongs to the original dataset owner. - NSynth(https://magenta.tensorflow.org/datasets/nsynth) under its license(https://creativecommons.org/licenses/by/4.0/). Copyright belongs to the original dataset owner. - Clotho-AQA(https://zenodo.org/records/6473207) under its license(mostly CreativeCommons with attribution). Copyright belongs to the original dataset owner. - MELD(https://affective-meld.github.io/) under its license. Copyright belongs to the original dataset owner. - SLURP(https://zenodo.org/records/4274930) under its license(Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license). Copyright belongs to the original dataset owner. - VoxCeleb1(https://www.robots.ox.ac.uk/~vgg/data/voxceleb/vox1.html) under its license(http://creativecommons.org/licenses/by-sa/4.0/). Copyright belongs to the original dataset owner. - LibriSpeech(https://www.openslr.org/12) under its license(CC BY 4.0). Copyright belongs to the original dataset owner. - CoVoST 2(https://huggingface.co/datasets/covost2) under its license(Creative Commons Attribution Non Commercial 4.0). Copyright belongs to the original dataset owner. - Fake-or-Real(https://bil.eecs.yorku.ca/datasets/) under its license. Copyright belongs to the original dataset owner. - VocalSound(https://github.com/YuanGongND/vocalsound) under its license(http://creativecommons.org/licenses/by-sa/4.0/). Copyright belongs to the original dataset owner. **Please contact us if you believe any data infringes upon your rights, and we will remove it.**