TOC-Bench - Event Count: leaderboard
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. 23 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Seed 2.0 Lite | 24.3 |
| 2 | GPT-5.4 Mini | 21.3 |
| 3 | Grok 4.3 | 19.7 |
| 4 | Kimi K2.6 | 19.3 |
| 5 | Gemini 3.1 Pro (Preview) | 18.7 |
| 6 | InternVL3-8B | 18.7 |
| 7 | MiniCPM-V-2.6 | 18.7 |
| 8 | GPT-5.5 | 18.3 |
| 9 | Qwen 3 VL 8B (Thinking) | 14 |
| 10 | GLM-5V Turbo | 13.7 |
| 11 | Gemini 3.1 Flash Lite (Preview) | 6 |
| 12 | MiMo-V2-Omni | 4.7 |
| 13 | Qwen 2.5 VL 7B | 0.7 |
Interactive version: theaggregate.ai/benchmark?slug=toc-bench-event-count · How It Works · Data refreshed daily, snapshot 2026-10-07.