SIS-Bench: leaderboard
Metric: Overall accuracy (%; 4,856 questions over 13 tasks on UAV spatial cognition and self-awareness at perception, memory and reasoning levels, question-weighted; four-option multiple choice on UAV videos (at most 32 frames), zero-shot, open models through vLLM with at most 128 generated tokens and proprietary models through their APIs; higher is better). Source: arxiv.org. Saturation forecast: Around September 2027. 26 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 71.6 |
| 2 | Kimi K2.5 | 71 |
| 3 | Seed 1.8 | 70.6 |
| 4 | Qwen 3.5 Plus | 70.1 |
| 5 | GPT-5.4 | 70 |
| 6 | Qwen 3 VL 8B (Thinking) | 64.7 |
| 7 | Qwen 3 VL 8B Instruct | 64.6 |
| 8 | Qwen 3 VL 4B Instruct | 62.5 |
| 9 | InternVL3.5-8B | 61.1 |
| 10 | Qwen 3 VL 30B A3B Instruct | 61 |
| 11 | GLM-4.1V-9B (Thinking) | 60.4 |
| 12 | Step3 VL 10B | 60.3 |
| 13 | InternVL3-14B | 59.1 |
| 14 | Qwen 2.5 VL 7B Instruct | 55.8 |
| 15 | Qwen 2 VL 7B Instruct | 53.2 |
Interactive version: theaggregate.ai/benchmark?slug=sis-bench · How It Works · Data refreshed daily, snapshot 2026-09-29.