VGenST-Bench - Direction Estimation: 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. Direction Estimation task: after a first-person walk through a room, give the direction of the starting point as a clock position (vista scale, egocentric, static). Accuracy in percent on multiple-choice questions under circular evaluation (right under every rotation of the options), 8 sampled frames; higher is better.
Metric: Direction Estimation (vista scale, egocentric, static) 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
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 76.4 | #138 |
| 2 | Gemini 3 Flash | 73.4 | #180 |
| 3 | Kimi K2.6 | 73.1 | #197 |
| 4 | Gemini 3.1 Flash Lite | 66.9 | #377 |
| 5 | GPT-5.4 Mini | 60 | #409 |
| 6 | GPT-5.4 Nano | 59.4 | #653 |
| 7 | Gemma 4 31B (IT) | 59.3 | #354 |
| 8 | InternVL3.5-30B-A3B | 59 | #1269 |
| 9 | GLM-4.6V-Flash | 56.6 | #1049 |
| 10 | Qwen 3.5 27B | 54.4 | #380 |
| 11 | Gemma 4 26B A4B (IT) | 54.3 | #563 |
| 12 | Qwen 3.5 9B | 52 | #838 |
| 13 | Qwen 3.5 4B | 50.4 | #1270 |
| 14 | InternVL3.5-8B | 46.6 | #1618 |
| 15 | InternVL3.5-4B | 41.4 | #1946 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=vgenst-bench-direction-estimation · How It Works · Data refreshed daily, snapshot 2026-11-22.