UAV-DualCog - Flight Behavior Recognition (Atomic): leaderboard
Metric: Atomic-level behavior accuracy (%; the atomic flight actions of the same 1,024 flight videos; 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 July 2028. 29 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.7 Plus (Non-reasoning) | 34 |
| 2 | Kimi K2.5 (Non-reasoning) | 33.6 |
| 3 | GLM-4.6V (Non-reasoning) | 32.5 |
| 4 | Kimi K2.6 (Non-reasoning) | 32.3 |
| 5 | Qwen 3.5 Plus (Non-reasoning) | 31.7 |
| 6 | Qwen 3.5 27B (Non-reasoning) | 31.3 |
| 7 | Qwen 3.5 9B (Non-reasoning) | 29.8 |
| 8 | Qwen 3.5 35B A3B (Non-reasoning) | 29.7 |
| 9 | Qwen 3.6 Plus (Non-reasoning) | 29 |
| 10 | Qwen 3.5 397B A17B (Non-reasoning) | 28.8 |
| 11 | MiMo-V2-Omni | 26.9 |
| 12 | Qwen 3.5 122B A10B (Non-reasoning) | 23.6 |
| 13 | Gemini 3.1 Flash Lite (Non-reasoning) | 23.3 |
| 14 | InternVL3.5-8B | 21.5 |
| 15 | Qwen 3.5 4B (Non-reasoning) | 7.5 |
Interactive version: theaggregate.ai/benchmark?slug=uav-dualcog-flight-behavior-recognition-atomic · How It Works · Data refreshed daily, snapshot 2026-09-29.