Datasets:

0x8badbeef
/

molmo-audio-avqa-interleaved

id string	text string	audio_bytes unknown	audio_sampling_rate int32	audio_codes list	modality_relation string	question_type string	video_name string	source string	split string
avqa-train-255	"Audio-visual question: What is the main source of sound in the video?\nOptions: aircraft; train; fi(...TRUNCATED)	"UklGRiTiBABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDiBACsFxQomBujHsAhFgYG+0/GUdIPsm+zQg6yKwMZWA4(...TRUNCATED)	16,000	[256,188,144,148,147,157,175,158,137,159,126,144,146,136,119,51,66,104,310,382,313,311,322,310,300,2(...TRUNCATED)	Both	Which	-KdBaGgzf4w_000027	avqa	train
avqa-train-220	Audio-visual question: What's calling? Options: chicken; crow; cock; lion Answer: crow	"UklGRiTiBABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDiBABHw1HW7/Gd9qEIzRAVHboWSPuv8RjohfEE69TwJA0(...TRUNCATED)	16,000	[51,120,55,99,91,50,115,95,108,92,110,110,93,91,112,52,121,39,129,78,103,80,97,89,35,107,64,86,116,5(...TRUNCATED)	Both	Which	-JC9g19P-vU_000035	avqa	train
avqa-train-201	"Audio-visual question: Where did the video take place?\nOptions: On the road; aquatic; street; By t(...TRUNCATED)	"UklGRlBmBABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YSxmBAA24j7cX9lW3z/xm/4fAWr1GfdIAh37VvIX5MD2i/g(...TRUNCATED)	16,000	[150,146,139,141,143,131,125,123,118,116,117,115,113,114,117,114,110,109,107,116,124,135,139,141,135(...TRUNCATED)	Both	Where	-II-7qJxKPc_000021	avqa	train
avqa-train-200	"Audio-visual question: What is the main source of sound in the video?\nOptions: toilet flush sound;(...TRUNCATED)	"UklGRiTiBABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDiBADT+X//p/jx8Svn7wPZO3QviP4m8Ej4Q+/g8qsQqhw(...TRUNCATED)	16,000	[147,136,142,145,135,144,139,148,147,133,158,148,145,128,144,144,138,137,145,135,149,146,140,146,145(...TRUNCATED)	Both	Happening	-IFRoPYS5cI_000001	avqa	train
avqa-train-228	"Audio-visual question: What is she doing?\nOptions: Rope skipping; rowing; Stamp one's foot; speech(...TRUNCATED)	"UklGRiTiBABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDiBABJAv/0RuU+4XzjNOd07y/5D/yH+8D5dfYL9Lj4NQB(...TRUNCATED)	16,000	[ 78, 70, 69, 51, 58, 91, 84, 99, 92, 90, 84, 86, 92, 64, 89, 88, 64, 107, 92, 84, 70, 63, 44, 77, 92, 70, 99, 103, 94, 102, 94, 70 ]	Both	Which	-J_Lm_eqp0U_000102	avqa	train
avqa-train-237	"Audio-visual question: What animal is eating grass?\nOptions: horse; pheasant; cattle; bird\nAnswer(...TRUNCATED)	"UklGRiTiBABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDiBAC2AIsEXwN++tz6SAcICb/6I/ZOAdQFjP5t/UYCBv9(...TRUNCATED)	16,000	[144,116,102,139,120,159,161,188,152,150,153,123,187,143,136,143,147,158,176,126,116,122,169,189,159(...TRUNCATED)	View	Happening	-JVYFW4VyKc_000117	avqa	train
avqa-train-267	"Audio-visual question: What does sound come from?\nOptions: train; automobile; The sound of the tru(...TRUNCATED)	"UklGRiTiBABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDiBABSHnQkQSi8JYwgHhyTF9IS6g5XEAURhRDQCVsPVRA(...TRUNCATED)	16,000	[100,104,97,99,92,89,97,89,100,102,94,102,105,95,88,93,81,77,77,84,86,74,77,104,99,92,79,113,90,82,1(...TRUNCATED)	Both	Come From	-LP9auWyYe8_000030	avqa	train
avqa-train-272	"Audio-visual question: What is the main source of sound in the video?\nOptions: The dog barked; che(...TRUNCATED)	"UklGRlDdBABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YSzdBABH/Iz+/wN3Af37Wv+TAzwAwv0KAVwBUv7M/zICkv8(...TRUNCATED)	16,000	[171,173,49,58,135,92,129,123,42,47,168,170,63,57,158,53,44,75,106,49,62,98,25,81,164,153,140,65,35,(...TRUNCATED)	Both	Come From	-Lei1usDQio_000020	avqa	train
avqa-train-218	"Audio-visual question: What are the people doing in the video?\nOptions: riding; diving; eat someth(...TRUNCATED)	"UklGRiTiBABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDiBACY+IL4/Pfs95v3xPfn9zj45/jO+d36ZvzD/Tz/sAD(...TRUNCATED)	16,000	[ 33, 33, 33, 36, 34, 31, 30, 34, 41, 38, 41, 33, 34, 29, 28, 31, 25, 26, 26, 24, 22, 24, 20, 23, 26, 31, 35, 37, 34, 41, 38, 43 ]	Both	Which	-JBRB2gGWDA_000004	avqa	train
avqa-train-305	"Audio-visual question: Where did the video take place?\nOptions: expressway; airport; square; Racet(...TRUNCATED)	"UklGRiTiBABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDiBACqF90XZh9RHXIXbAiQ+0sKEA75BGQF4/3Z+BL3fPz(...TRUNCATED)	16,000	[118,129,141,133,124,121,109,94,105,114,114,105,137,108,113,132,117,121,125,122,107,94,92,106,106,10(...TRUNCATED)	Both	Where	-Nis-cumBYc_000250	avqa	train

End of preview. Expand in Data Studio

Molmo-Audio AVQA Interleaved

Prepared interleaved text + image + audio dataset for training cross-modal models such as Molmo-Audio (Molmo vision-language + SLAP audio).

Source

Derived from juyil/AVQA-videos (AVQA annotations + VGGSound clips). Original rights remain with AVQA / VGGSound authors and YouTube uploaders. Research / non-commercial use only.

Why this dataset

There is no high-quality open HF dataset that already ships Molmo-style interleaved image+audio+text ready for NTP training. AVQA is the strongest practical base: aligned soundtrack + visuals + language supervision, with media bundled on the Hub (no YouTube scraping).

Format

Each row:

field	description
`text`	Instruction-style AV QA string (question, options, answer)
`image`	RGB keyframe (~1.5s) from the clip
`audio_bytes`	Mono 16 kHz WAV bytes of the clip soundtrack
`audio_sampling_rate`	Always `16000`
`audio_codes`	32 discrete pseudo-VQ codes (codebook 1024) for discrete-audio NTP bootstrap
`modality_relation`	AVQA label: `Both` / `Sound` / `View`
`question_type`	AVQA question category
`video_name`	Source clip id
`source`	Always `avqa`

Splits

train: 7071 examples
validation: 2957 examples

Repo: 0x8badbeef/molmo-audio-avqa-interleaved

Citation

Please cite AVQA and VGGSound when using this data:

@inproceedings{yang2022avqa,
  title={AVQA: A Dataset for Audio-Visual Question Answering on Videos},
  author={Yang, Pinci and Wang, Xin and Duan, Xuguang and Chen, Hong and
           Hou, Runze and Jin, Cong and Zhu, Wenwu},
  booktitle={ACM MM},
  year={2022}
}
@inproceedings{chen2020vggsound,
  title={VGGSound: A Large-scale Audio-Visual Dataset},
  author={Chen, Honglie and Xie, Weidi and Vedaldi, Andrea and Zisserman, Andrew},
  booktitle={ICASSP},
  year={2020}
}

Downloads last month: 73