UAV-DualCog - Landmark-Relative Direction: leaderboard
Metric: Answer accuracy (%; 1,024 questions asking where a landmark lies relative to the UAV forward direction; structured JSON answers on renders of simulated AerialVLN scenes, the discrete answer scored whatever the predicted box or interval; instant mode with explicit thinking disabled where a model allows it; higher is better). Source: arxiv.org. Saturation forecast: Around 2028. 36 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 (Non-reasoning) | 65.4 |
| 2 | Qwen 3.5 27B (Non-reasoning) | 57.8 |
| 3 | GPT-5.3 Instant | 56.5 |
| 4 | Qwen 3.7 Plus (Non-reasoning) | 55.4 |
| 5 | Qwen 3.5 9B (Non-reasoning) | 53.2 |
| 6 | Qwen 3.5 35B A3B (Non-reasoning) | 53.2 |
| 7 | Qwen 3.5 122B A10B (Non-reasoning) | 49.7 |
| 8 | Claude Sonnet 4.6 | 48.5 |
| 9 | Qwen 3.6 Plus (Non-reasoning) | 48.4 |
| 10 | Qwen 3.5 397B A17B (Non-reasoning) | 47.9 |
| 11 | Qwen 3.5 Plus (Non-reasoning) | 47.6 |
| 12 | Gemini 3.1 Flash Lite (Non-reasoning) | 47.6 |
| 13 | Qwen 3.5 4B (Non-reasoning) | 47.5 |
| 14 | Kimi K2.6 (Non-reasoning) | 42.8 |
| 15 | Kimi K2.5 (Non-reasoning) | 39.4 |
Interactive version: theaggregate.ai/benchmark?slug=uav-dualcog-landmark-relative-direction · How It Works · Data refreshed daily, snapshot 2026-09-29.