TOC-Bench - Event Existence: leaderboard

The event existence 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: whether a described object event happens in the video. Accuracy in percent; higher is better.

Metric: Event Existence 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 July 2027. 21 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.568.9#54
2Seed 2.0 Lite57.5#300
3Kimi K2.656.6#192
4Gemini 3.1 Pro (Preview)56.2#99
5GLM-5V Turbo52.1#360
6Video-LLaVA-7B51.1#3879
7LLaVA-Video-72B49.3#1193
8Grok 4.348.9#342
9Qwen 2.5 VL 72B47.5#781
10MiMo-V2-Omni47.5#465
11GPT-5.4 Mini46.1#406
12VideoChat-R1.5-7B45.2#1254
13InternVL3-8B42.5#1782
14Qwen 2.5 VL 7B40.6#1862
15Gemini 3.1 Flash Lite (Preview)37.9#417

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