AirGroundBench - Counterfactual Reasoning: leaderboard
AirGroundBench probes spatial intelligence for heterogeneous air-ground (UAV plus UGV) embodied collaboration, with nine multiple-choice VQA tasks in four capability dimensions plus a vision-language navigation task, evaluated zero-shot under synchronized dual views. This board reports Accuracy (%) on the counterfactual spatial-reasoning VQA task; higher is better.
Metric: Accuracy (%) on the counterfactual spatial-reasoning VQA task; higher is better. Source: arxiv.org. Saturation forecast: Around November 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 Plus | 48.57 | #232 |
| 2 | Qwen 3.5 397B A17B | 45.71 | #246 |
| 3 | GPT-5.2 | 44.93 | #176 |
| 4 | Seed 2.0 Pro | 44.62 | #211 |
| 5 | Claude Sonnet 4.6 | 44.35 | #161 |
| 6 | Gemini 3.1 Pro (Preview) | 43.42 | #94 |
| 7 | Qwen 3.5 27B | 43.25 | #359 |
| 8 | Kimi K2.5 | 40.3 | #256 |
| 9 | GLM-4.6V | 38.46 | #701 |
| 10 | ERNIE 5.0 | 35.82 | #456 |
| 11 | Qwen 3 VL 235B A22B | 35.38 | #479 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=airgroundbench-counterfactual-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-13.