TOC-Bench - Conditional State: leaderboard
Metric: Conditional State 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 2030. 23 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Kimi K2.6 | 56.5 |
| 2 | Grok 4.3 | 51.8 |
| 3 | Gemini 3.1 Pro (Preview) | 50 |
| 4 | Qwen 2.5 VL 7B | 43.9 |
| 5 | Seed 2.0 Lite | 43.5 |
| 6 | MiMo-V2-Omni | 42.1 |
| 7 | GPT-5.5 | 41.7 |
| 8 | Qwen 3 VL 8B (Thinking) | 40.6 |
| 9 | Gemini 3.1 Flash Lite (Preview) | 39.6 |
| 10 | GLM-5V Turbo | 36.7 |
| 11 | GPT-5.4 Mini | 36.3 |
| 12 | InternVL3-8B | 31.7 |
| 13 | MiniCPM-V-2.6 | 25.2 |
Interactive version: theaggregate.ai/benchmark?slug=toc-bench-conditional-state · How It Works · Data refreshed daily, snapshot 2026-10-07.