VBVR-Bench

VBVR-Bench evaluates video generation models on visual behavior and video reasoning across in-domain and out-of-domain abstraction, knowledge, perception, spatial, and transition categories.

13rows
overallprimary metric
2026-05-06sampled

Metadata

Metrics

Overall, Overall(In-Domain), Overall(Out-of-Domain), Abst.(All), Know.(All), Perc.(All), Spat.(All), Trans.(All), Abst.(ID), Know.(ID), Perc.(ID), Spat.(ID), Trans.(ID), Abst.(OOD), Know.(OOD), Perc.(OOD), Spat.(OOD), Trans.(OOD)

Latest Results

Rank Subject Overall Model Match Provenance Sampled
1 Human 0.97 Imported 2026-05-06
2 VBVR-Wan2.2 0.68 Imported 2026-05-06
3 VBVR-Wan2.1 0.59 Imported 2026-05-06
4 Sora 2 0.55 Imported 2026-05-06
5 Seedance 2.0 0.54 Imported 2026-05-06
6 VBVR-LTX2.3 0.52 Imported 2026-05-06
7 Veo 3.1 0.48 Imported 2026-05-06
8 Runway Gen-4 Turbo 0.40 Imported 2026-05-06
9 Wan2.2-I2V-A14B 0.37 Imported 2026-05-06
10 Kling 2.6 0.37 Imported 2026-05-06
11 LTX-2 0.31 Imported 2026-05-06
12 CogVideoX1.5-5B-I2V 0.27 Imported 2026-05-06
13 HunyuanVideo-I2V 0.27 Imported 2026-05-06