VGI-Bench: Probing Visual Intelligence in Video Generation Models
VGI-Bench evaluates video generation models via 27 visual reasoning tasks, finding top models achieve only 51% accuracy with limited self-correction.
Published Aug 20, 2026▲ 336 on Hugging FaceCode ★ 14arXiv ↗
Only vote on papers you've read. Sign in with GitHub to vote.
VGI-Bench delivers a rigorous 810-instance taxonomy of evolving visual reasoning that exposes limited self-correction, though its 51% ceiling and synthetic failure modes leave clinical relevance and task novelty contested.
Abstract
Recent studies suggest that video generation models can exhibit certain forms of zero-shot visual reasoning through generated frames. Yet reliable evaluation remains challenging: benchmarks should adopt inputs aligned with the visual priors of current video models, require valid evolving processes rather than only plausible final states, and calibrate task difficulty to remain challenging yet partly feasible. To this end, we introduce VGI-bench, containing 27 tasks and 810 instances, organized by a two-level taxonomy of task domains and skill tags for fine-grained evaluation of visual reasoning capabilities of video generation models. Our evaluations show that current generative systems can solve a subset of visually grounded reasoning tasks, but remain far from reliable, with even the strongest model, Seedance 2.0, achieving only 51.0% under our evaluation criteria. Our analysis further explore the output failure modes, input condition sensitivity, performance transfer boundary from synthetic fine-tuning, and internal denoising perspective revealing limited self-correction, where later steps mainly refine early hypotheses rather than correct reasoning errors. We hope VGI-bench will help stimulate the development of next-generation video generation models. Website: https://hexuan21.github.io/VGI-Bench/