TOC-Bench - Temporal Location: leaderboard

The temporal location 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: when in the video an object event happens (beginning, early, late or end). Accuracy in percent; higher is better.

Metric: Temporal Location 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 October 2028. 21 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.548.6#56
2Kimi K2.647.4#197
3Gemini 3.1 Pro (Preview)45#102
4Grok 4.342.9#351
5Seed 2.0 Lite40.9#306
6Qwen 2.5 VL 72B38.2#782
7Gemini 3.1 Flash Lite (Preview)38#419
8Qwen 3 VL 8B (Thinking)34.9#924
9GLM-5V Turbo34.8#365
10GPT-5.4 Mini34.1#409
11InternVL3-8B32#1795
12MiMo-V2-Omni30.3#465
13VideoChat-R1.5-7B29.3#1260
14LLaVA-Video-7B28.6#2303
15VideoLLaMA3-7B26.4#2064

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=toc-bench-temporal-location · How It Works · Data refreshed daily, snapshot 2026-11-22.