ERGeoBench - Spatial Awareness: leaderboard

Spatial Awareness diagnostics on ERGeoBench (2026): yaw, position, depth, distance, relative position and cross-view questions in egocentric street views, averaged over seven sub-tasks. Accuracy in percent; higher is better.

Metric: Accuracy (%) on the spatial awareness diagnostic questions: yaw estimation, 2D positioning, depth, distance, left/right, front/back and cross-view questions judged by explicit matching rules, averaged over the seven sub-tasks; unified structured prompt with egocentric observations, action history and in-context examples, temperature 0.1, first valid JSON object parsed (invalid outputs count as failures); higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 9 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.0 Flash52.55#723
2Qwen 3 VL 8B49.67#917
3Qwen-VL-Max48.93#1400
4GPT-4o43.86#751
5InternVL3-8B43.61#1782
6Qwen 2.5 VL 7B42.91#1862
7InternVL2.5-8B42.31#2370
8Gemini 2.5 Pro40.19#276
9Gemini 3 Flash37.03#175

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

Interactive version: theaggregate.ai/benchmark?slug=ergeobench-spatial-awareness · How It Works · Data refreshed daily, snapshot 2026-11-21.