CrashSight - Temporal Sequence: leaderboard

Temporal sequence questions (event order and crash phase, unanswerable from one frame) of CrashSight (2026), a multiple-choice VQA benchmark on real roadside-camera traffic crash videos answered zero-shot from four frames. Accuracy in percent; higher is better.

Metric: Accuracy (%) on temporal-sequence questions (event order and phase identification, unanswerable from one frame), four-option multiple-choice VQA on held-out roadside-camera crash videos of CrashSight-VQA (4 frames at 1 fps, zero-shot, deterministic decoding, exact letter match); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1InternVL3-8B85.7#1795
2InternVL3-2B81#3318
3Qwen 2.5 VL 7B Instruct66.7#1919
4Qwen 2.5 VL 3B Instruct64.3#2720
5LLaVA-OneVision-0.5B52.4#7977
6LLaVA-NeXT-Video-7B42.9#5783

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

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