EgoDyn-Bench - Temporal F1: leaderboard

EgoDyn-Bench (2026) tests whether vision-language models understand the motion of the ego vehicle in driving video, through multiple-choice questions on speed, acceleration and turning answered from video alone. This board is macro-F1 in percent on the temporal questions about how the motion changes during a clip; higher is better.

Metric: Macro-F1 (%) of temporal ego-motion questions (how the vehicle's motion changes over the clip) from driving video alone; higher is better. Source: arxiv.org. Saturation forecast: Around October 2027. 16 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro (Preview)45.2#119
2GPT-5.138.7#240
3Qwen 3 VL 2B37.5#2134
4Cosmos-Reason2-8B36.7#755
5Gemini 2.0 Flash35.1#723
6Qwen 3 VL 4B35.1#1151
7Claude Sonnet 4.534.9#261
8InternVL3.5-8B33.5#1604
9Qwen 3 VL 8B32.7#917
10RoboTron-Drive31.3
11ImpromptuVLA (EgoDyn-Bench checkpoint unspecified)29.2
12InternVL3.5-4B28.7#1930
13CamReasoner-8B28.6
14InternVL3.5-38B27#1021
15InternVL3-2B26.4#3290

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=egodyn-bench-temporal-f1 · How It Works · Data refreshed daily, snapshot 2026-11-21.