SIS-Bench - Path Planning: leaderboard
SIS-Bench (2026) tests multimodal LLMs on UAV self-awareness and spatial cognition with four-option questions about drone videos. The path planning task asks which action sequence reaches a target from a start point in a long flight; accuracy in percent, higher is better.
Metric: Accuracy (%; 272 path planning questions; 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 February 2027. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 58.8 | #123 |
| 2 | Qwen 3.5 Plus | 58.5 | #232 |
| 3 | Kimi K2.5 | 57 | #256 |
| 4 | Seed 1.8 | 54 | #242 |
| 5 | Gemini 3 Flash (Preview) | 53.7 | #144 |
| 6 | Qwen 3 VL 8B Instruct | 48.9 | #981 |
| 7 | Qwen 3 VL 30B A3B Instruct | 48.9 | #871 |
| 8 | Qwen 3 VL 8B (Thinking) | 46.3 | #887 |
| 9 | GLM-4.1V-9B (Thinking) | 43.4 | #1060 (GLM-4.1V-9B) |
| 10 | Qwen 3 VL 4B Instruct | 36.4 | #1279 |
| 11 | Qwen 2.5 VL 7B Instruct | 31.6 | #1833 |
| 12 | InternVL3-14B | 31.6 | #1182 |
| 13 | Qwen 2 VL 7B Instruct | 22.1 | #2553 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=sis-bench-path-planning · How It Works · Data refreshed daily, snapshot 2026-11-13.