LinkS2Bench - Scene Association: leaderboard
LinkS2Bench (LinkS²Bench, 2026) task on drone videos paired with satellite imagery: multiple-choice accuracy (%) on which satellite region matches a video segment; higher is better.
Metric: Scene Association: multiple-choice accuracy (%) on which satellite region matches a video segment on the full LinkS2Bench (17,903 questions over 1,500 UAV videos paired with satellite images), temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 18 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 397B A17B | 73.3 | #264 |
| 2 | GPT-5.4 | 66.8 | #135 |
| 3 | Seed 1.8 | 66.7 | #252 |
| 4 | Gemini 3.1 Pro (Preview) | 57.3 | #99 |
| 5 | GLM-4.6V-Flash | 57.1 | #1040 |
| 6 | Qwen 3.5 35B A3B | 54.2 | #492 |
| 7 | Kimi K2.5 | 51 | #266 |
| 8 | Gemini-3.1-Flash [LinkS2Bench checkpoint unspecified] | 47 | |
| 9 | Claude Sonnet 4.6 | 43.3 | #173 |
| 10 | GPT-5.4 Mini | 39.3 | #406 |
| 11 | LLaVA-OneVision-1.5-8B | 24.4 | #1864 |
| 12 | Qwen 3.5 9B | 19 | #828 |
| 13 | LLaVA-OneVision-1.5-4B | 16.7 | #1855 |
| 14 | Ministral 3 3B | 11.9 | #3504 |
| 15 | Llama 4 Scout | 10.4 | #1818 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=links2bench-scene-association · How It Works · Data refreshed daily, snapshot 2026-11-21.