VGenST-Bench - Quantity Change Tracking: 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. Quantity Change Tracking task: track the item counts on a workspace as items are added or removed (figural 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: Quantity Change Tracking (figural 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 3.1 Flash Lite71.6#371
2Gemini 3 Flash71.4#175
3GPT-5.471#135
4Gemma 4 31B (IT)66.9#350
5Gemma 4 26B A4B (IT)61.9#556
6Qwen 3.5 27B57#375
7InternVL3.5-30B-A3B55.7#1267
8Qwen 3.5 9B54.4#828
9Qwen 3.5 4B51.6#1241
10GPT-5.4 Nano48.2#651
11GPT-5.4 Mini46.3#406
12Kimi K2.645#192
13InternVL3.5-8B42#1604
14GLM-4.6V-Flash40#1040
15InternVL3.5-4B38.3#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-quantity-change-tracking · How It Works · Data refreshed daily, snapshot 2026-11-21.