TOC-Bench - Relative Spatial Change: leaderboard

The relative spatial change questions of TOC-Bench (2026), video questions grounded in per-frame object tracks and kept only when not answerable from text alone, single frames or shuffled frames: which way an object moves across the frame, or whether it stays in place. Accuracy in percent; higher is better.

Metric: Relative Spatial Change accuracy (%) on TOC-Bench (2,323 human-verified questions over 1,951 videos from Charades, Perception Test, OVIS and MOSE, grounded in per-frame object tracks and kept only when not answerable from text alone, single frames or shuffled frames), mostly 32 uniformly sampled frames, deterministic scoring of multiple-choice, statement-pair, numeric and ordering formats; higher is better. Source: arxiv.org. Saturation forecast: Around 2028. 21 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Grok 4.346.6#342
2Gemini 3.1 Pro (Preview)43.8#99
3GPT-5.543.8#54
4Kimi K2.643.8#192
5Seed 2.0 Lite38.2#300
6GLM-5V Turbo33.1#360
7Qwen 3 VL 8B (Thinking)30.3#910
8GPT-5.4 Mini28.7#406
9Gemini 3.1 Flash Lite (Preview)28.7#417
10Qwen 2.5 VL 72B27#781
11VideoLLaMA3-7B27#2050
12MiMo-V2-Omni27#465
13LLaVA-Video-7B27#2276
14InternVL3-8B26.4#1782
15Qwen 2.5 VL 7B25.8#1862

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

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