MMTEB: Massive Multilingual Text Embedding Benchmark
Paper • 2502.13595 • Published • 50
Datasets:
id string | text string |
|---|---|
0 | a video of someone playing accordion |
1 | a video of someone playing acoustic guitar |
2 | a video of someone playing bagpipe |
3 | a video of someone playing banjo |
4 | a video of someone playing bassoon |
5 | a video of someone playing cello |
6 | a video of someone playing clarinet |
7 | a video of someone playing congas |
8 | a video of someone playing drum |
9 | a video of someone playing electric bass |
10 | a video of someone playing erhu |
11 | a video of someone playing flute |
12 | a video of someone playing guzheng |
13 | a video of someone playing piano |
14 | a video of someone playing pipa |
15 | a video of someone playing saxophone |
16 | a video of someone playing suona |
17 | a video of someone playing trumpet |
18 | a video of someone playing tuba |
19 | a video of someone playing ukulele |
20 | a video of someone playing violin |
21 | a video of someone playing xylophone |
MUSIC-AVQA classification dataset containing 22 instrument categories. Given a video and audio of someone playing an instrument, the goal is to predict the instrument type. This zero-shot variant uses both video and audio modalities. Filtered the test split to rows with a 22-class instrument answer (~1,706 examples).
| Task category | VideoZeroshotClassification (video+audio-to-text) |
| Domains | Music |
| Reference | Proceedings of the IEEE/CVF conference on computer vision and pattern recognition |
Source datasets:
How to evaluate on this task
You can evaluate an embedding model on this dataset using the following code:
import mteb
task = mteb.get_task("MusicAVQACLSAudioVideoZeroShot")
model = mteb.get_model(YOUR_MODEL)
mteb.evaluate(model, task)
To learn more about how to run models on mteb task check out the GitHub repository.
Citation
If you use this dataset, please cite the dataset as well as mteb, as this dataset likely includes additional processing as a part of the MMTEB Contribution.
@inproceedings{li2022learning,
author = {Li, Guangyao and Wei, Yake and Tian, Yapeng and Xu, Chenliang and Wen, Ji-Rong and Hu, Di},
booktitle = {Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},
pages = {19108--19118},
title = {Learning to answer questions in dynamic audio-visual scenarios},
year = {2022},
}
@article{enevoldsen2025mmtebmassivemultilingualtext,
title={MMTEB: Massive Multilingual Text Embedding Benchmark},
author={Kenneth Enevoldsen and Isaac Chung and Imene Kerboua and Márton Kardos and Ashwin Mathur and David Stap and Jay Gala and Wissam Siblini and Dominik Krzemiński and Genta Indra Winata and Saba Sturua and Saiteja Utpala and Mathieu Ciancone and Marion Schaeffer and Gabriel Sequeira and Diganta Misra and Shreeya Dhakal and Jonathan Rystrøm and Roman Solomatin and Ömer Çağatan and Akash Kundu and Martin Bernstorff and Shitao Xiao and Akshita Sukhlecha and Bhavish Pahwa and Rafał Poświata and Kranthi Kiran GV and Shawon Ashraf and Daniel Auras and Björn Plüster and Jan Philipp Harries and Loïc Magne and Isabelle Mohr and Mariya Hendriksen and Dawei Zhu and Hippolyte Gisserot-Boukhlef and Tom Aarsen and Jan Kostkan and Konrad Wojtasik and Taemin Lee and Marek Šuppa and Crystina Zhang and Roberta Rocca and Mohammed Hamdy and Andrianos Michail and John Yang and Manuel Faysse and Aleksei Vatolin and Nandan Thakur and Manan Dey and Dipam Vasani and Pranjal Chitale and Simone Tedeschi and Nguyen Tai and Artem Snegirev and Michael Günther and Mengzhou Xia and Weijia Shi and Xing Han Lù and Jordan Clive and Gayatri Krishnakumar and Anna Maksimova and Silvan Wehrli and Maria Tikhonova and Henil Panchal and Aleksandr Abramov and Malte Ostendorff and Zheng Liu and Simon Clematide and Lester James Miranda and Alena Fenogenova and Guangyu Song and Ruqiya Bin Safi and Wen-Ding Li and Alessia Borghini and Federico Cassano and Hongjin Su and Jimmy Lin and Howard Yen and Lasse Hansen and Sara Hooker and Chenghao Xiao and Vaibhav Adlakha and Orion Weller and Siva Reddy and Niklas Muennighoff},
publisher = {arXiv},
journal={arXiv preprint arXiv:2502.13595},
year={2025},
url={https://arxiv.org/abs/2502.13595},
doi = {10.48550/arXiv.2502.13595},
}
@article{muennighoff2022mteb,
author = {Muennighoff, Niklas and Tazi, Nouamane and Magne, Loïc and Reimers, Nils},
title = {MTEB: Massive Text Embedding Benchmark},
publisher = {arXiv},
journal={arXiv preprint arXiv:2210.07316},
year = {2022}
url = {https://arxiv.org/abs/2210.07316},
doi = {10.48550/ARXIV.2210.07316},
}
Dataset Statistics
Dataset Statistics
The following code contains the descriptive statistics from the task. These can also be obtained using:
import mteb
task = mteb.get_task("MusicAVQACLSAudioVideoZeroShot")
desc_stats = task.metadata.descriptive_stats
{
"test": {
"num_samples": 1706,
"text_statistics": null,
"image_statistics": null,
"audio_statistics": {
"total_duration_seconds": 101300.15399999925,
"min_duration_seconds": 14.65575,
"average_duration_seconds": 59.378753810081626,
"max_duration_seconds": 60.8925,
"unique_audios": 1693,
"average_sampling_rate": 16000.0,
"sampling_rates": {
"16000": 1706
}
},
"video_statistics": {
"total_duration_seconds": 101285.83323299981,
"total_frames": 6077150,
"min_width": 160,
"average_width": 695.4431418522861,
"max_width": 1920,
"min_height": 120,
"average_height": 393.4841735052755,
"max_height": 720,
"min_duration_seconds": 14.633333,
"average_duration_seconds": 59.37035945662357,
"max_duration_seconds": 60.866667,
"unique_videos": 1706,
"average_fps": 60.0,
"fps": {
"60": 1706
},
"min_resolution": [
160,
120
],
"average_resolution": [
695.4431418522861,
393.4841735052755
],
"max_resolution": [
1920,
720
],
"resolutions": {
"640x360": 1139,
"480x270": 10,
"1280x720": 158,
"576x324": 1,
"320x240": 36,
"480x360": 149,
"636x360": 9,
"640x338": 2,
"540x360": 14,
"568x320": 18,
"210x360": 2,
"308x360": 1,
"640x356": 1,
"640x358": 2,
"192x144": 1,
"640x320": 4,
"640x342": 2,
"1280x716": 1,
"448x336": 1,
"576x360": 1,
"480x272": 4,
"638x360": 3,
"484x272": 1,
"202x360": 10,
"542x360": 1,
"160x120": 1,
"406x720": 1,
"640x352": 10,
"204x360": 1,
"288x360": 2,
"470x360": 1,
"640x268": 1,
"640x354": 1,
"360x360": 3,
"640x272": 3,
"450x360": 4,
"490x360": 2,
"634x360": 2,
"320x180": 1,
"1280x622": 1,
"630x360": 2,
"628x360": 1,
"492x360": 1,
"478x360": 1,
"480x356": 1,
"270x360": 1,
"640x270": 3,
"626x360": 3,
"400x226": 1,
"1280x360": 23,
"322x240": 1,
"824x720": 1,
"632x360": 1,
"400x224": 2,
"1920x720": 18,
"1024x360": 2,
"296x224": 1,
"384x288": 3,
"1180x360": 1,
"1600x720": 2,
"1760x720": 2,
"800x360": 1,
"1040x360": 1,
"480x320": 3,
"1120x360": 6,
"350x240": 1,
"640x350": 1,
"352x262": 1,
"640x326": 1,
"208x360": 1,
"480x340": 1,
"1132x360": 1,
"960x360": 1,
"512x288": 1,
"1904x720": 1,
"320x214": 1,
"406x360": 1,
"472x360": 1,
"480x352": 2,
"960x720": 1,
"240x180": 1,
"608x360": 1,
"1090x360": 1
}
},
"label_statistics": {
"min_labels_per_text": 1,
"average_label_per_text": 1.0,
"max_labels_per_text": 1,
"unique_labels": 22,
"labels": {
"20": {
"count": 205
},
"1": {
"count": 117
},
"12": {
"count": 59
},
"15": {
"count": 97
},
"13": {
"count": 158
},
"11": {
"count": 98
},
"21": {
"count": 38
},
"3": {
"count": 49
},
"0": {
"count": 104
},
"18": {
"count": 72
},
"10": {
"count": 56
},
"5": {
"count": 130
},
"4": {
"count": 72
},
"17": {
"count": 74
},
"16": {
"count": 24
},
"7": {
"count": 21
},
"19": {
"count": 88
},
"14": {
"count": 46
},
"9": {
"count": 28
},
"8": {
"count": 53
},
"6": {
"count": 81
},
"2": {
"count": 36
}
}
},
"candidates_labels_text_statistics": {
"total_text_length": 745,
"min_text_length": 31,
"average_text_length": 33.86363636363637,
"max_text_length": 42,
"unique_texts": 22
}
}
}
This dataset card was automatically generated using MTEB
- Downloads last month
- -