TOC-Bench - Event Ordering: leaderboard

The event ordering 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: the chronological order of several events. Accuracy in percent; higher is better.

Metric: Event Ordering 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 2029. 21 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.524.8#54
2Kimi K2.622.9#192
3Seed 2.0 Lite21.9#300
4Gemini 3.1 Pro (Preview)17.1#99
5Grok 4.317.1#342
6GLM-5V Turbo17.1#360
7Gemini 3.1 Flash Lite (Preview)16.2#417
8VideoLLaMA3-7B16.2#2050
9Qwen 2.5 VL 72B15.2#781
10Qwen 2.5 VL 7B14.3#1862
11Video-R114.3#1379
12LLaVA-Video-72B14.3#1193
13GPT-5.4 Mini11.4#406
14Qwen 3 VL 8B (Thinking)10.5#910
15MiniCPM-V-2.69.5#3267

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-event-ordering · How It Works · Data refreshed daily, snapshot 2026-11-21.