TOC-Bench - Event Count: leaderboard

The event count 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: how many times an object event repeats. Accuracy in percent; higher is better.

Metric: Event Count 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: Not forecast. 21 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1VideoLLaMA3-7B33.7#2050
2Video-ChatGPT30.7#6929
3LLaVA-Video-7B26.3#2276
4Seed 2.0 Lite24.3#300
5GPT-5.4 Mini21.3#406
6Grok 4.319.7#342
7Kimi K2.619.3#192
8Gemini 3.1 Pro (Preview)18.7#99
9InternVL3-8B18.7#1782
10MiniCPM-V-2.618.7#3267
11GPT-5.518.3#54
12LLaVA-Video-72B15#1193
13Qwen 3 VL 8B (Thinking)14#910
14GLM-5V Turbo13.7#360
15VideoChat-R1.5-7B7#1254

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