DashBench - Q4 Event Detection: leaderboard

DashBench evaluates video understanding on 1,709 driving clips. Five explanation prompts and binary event detection are scored separately; mean agreement excludes the binary task. The 823-clip corner-case-window subset and training-free decoder variants retain their own conditions. Q5 asks for definition-guided corner-case identification and description. This board reports binary event-detection accuracy (%). Higher is better.

Metric: Binary event-detection accuracy (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked. Updated 14 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-597.83#167
2Gemini 3 Flash94.85#163
3MiniCPM-V-4.587.53#1390
4InternVL3.5-8BEvaluation run: DashBench SECOND81.33#1558
5InternVL3.5-38B77.65#986
6InternVL3.5-241B76.36#625
7Qwen 3 VL 8B74.96#899
8Qwen 3 VL 235B A22B72.03#484
9Qwen 3 VL 32B71.04#609
10Ovis2.5-9B64.47#1236
11VITA-1.561.32#3196
12InternLM-XComposer2.554.44#3981
13LLaVA-OV-1.5-8B54.13#4771
14InternVideo2.5-Chat-8B54#2017
15GLM-4.1V-9B (Thinking)35.34

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=dashbench-q4-event-detection · How It Works · Data refreshed daily, snapshot 2026-11-16.