SciWorldBench: video models score 1.4% to 24.8% on paired science experiments
A new benchmark of 222 experiment pairs finds ten text-and-image-to-video models rarely get both outcomes of a science experiment right.
Articles that report published prices, limits or benchmark figures.
A new benchmark of 222 experiment pairs finds ten text-and-image-to-video models rarely get both outcomes of a science experiment right.
An arXiv paper presents MemCrafter, an agentic prompt-refinement method for memorable text-to-video, and reports gains over Seedance-2.5 and other baselines.
A paper's visual harness for coding agents scores 58.6% on a 100-task video editing benchmark; we review ablations, MIT licence and API costs.