Research
Code2Video-Compass benchmarks code-based video generation across T2V, I2V and V2V
The benchmark has 2,104 tasks across 10 categories and 24 subtypes, scored with 12 objective and five subjective metrics. Evaluation of 15 MLLMs found performance declines as complexity rises.
Every outlet that ran it 1 article
Hugging Face PapersResearchLead