Datasets:
Tasks:
Video Classification
Formats:
json
Languages:
English
Size:
100K - 1M
ArXiv:
Tags:
reward-model
video-generation
evaluation
preference-learning
temporal-reasoning
synthetic-data
License:
Commit History
4 days ago
-
Add train and validation splits; remove the conflicted and vq_steps_16 subsets
Verified signatureTraining data merged from stellalisy/IsoVRBench-corpus: 160,000 train and 4,000 validation pairs, 234,075 clips in 66 tar shards under train/. The 8,000 benchmark pairs are the test split. conflicted.jsonl, vq_steps_16.jsonl and the 6,000 clips only they used are removed.b43943d