VGenST-Bench - Relative Velocity Identification: leaderboard

VGenST-Bench (2026) tests spatio-temporal reasoning of multimodal LLMs on short videos synthesized from validated scene graphs with text-to-image and image-to-video models. Relative Velocity Identification task: identify the motion of an entity relative to the moving observer (environmental scale, egocentric, dynamic). Accuracy in percent on multiple-choice questions under circular evaluation (right under every rotation of the options), 8 sampled frames; higher is better.

Metric: Relative Velocity Identification (environmental scale, egocentric, dynamic) accuracy (%) on the base multiple-choice questions of VGenST-Bench under circular evaluation (correct only if right under every cyclic permutation of the options), macro-averaged over the task's QA types, 8 uniformly sampled frames from about 100 synthesized videos per task (generated with text-to-image and image-to-video models from validated scene graphs); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash91.1#175
2GPT-5.481.9#135
3Gemini 3.1 Flash Lite77.2#371
4GPT-5.4 Mini73.8#406
5Gemma 4 31B (IT)67.8#350
6InternVL3.5-30B-A3B66.8#1267
7Kimi K2.664.2#192
8Gemma 4 26B A4B (IT)62.8#556
9GLM-4.6V-Flash59.3#1040
10Qwen 3.5 27B58.2#375
11GPT-5.4 Nano57.9#651
12Qwen 3.5 9B56.3#828
13InternVL3.5-8B53.8#1604
14Qwen 3.5 4B53.2#1241
15InternVL3.5-4B51.2#1930

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=vgenst-bench-relative-velocity-identification · How It Works · Data refreshed daily, snapshot 2026-11-21.