VGenST-Bench - Causal Mapping: 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. Causal Mapping task: link each triggering event to the state change it causes (figural scale, exocentric, 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: Causal Mapping (figural scale, exocentric, 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 Flash82.2#175
2GPT-5.479#135
3Kimi K2.676.7#192
4Gemini 3.1 Flash Lite76.2#371
5GPT-5.4 Mini72.5#406
6GLM-4.6V-Flash69.2#1040
7Gemma 4 31B (IT)67.5#350
8Qwen 3.5 27B65.6#375
9Qwen 3.5 9B63.3#828
10Gemma 4 26B A4B (IT)62.5#556
11InternVL3.5-30B-A3B62.3#1267
12Qwen 3.5 4B61.4#1241
13GPT-5.4 Nano52#651
14InternVL3.5-8B48.7#1604
15InternVL3.5-4B46.9#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-causal-mapping · How It Works · Data refreshed daily, snapshot 2026-11-21.