DK-VLM - Six-benchmark Mean: leaderboard
DK-VLM v1; LoRA/geometric attention fine-tunes remain distinct from Qwen2.5VL7BInstruct and Qwen3VL8BInstruct backbones. Frozen VGGT-Omega 1B frontend; official shared six-benchmark GeoThinker evaluation settings. Comparison baseline scores are source-reported, starred GeoThinker values reproduced by authors; no automatic primary-board fusion. Table2 six spatial datasets; VSI overall already retained once from Table1, not duplicated here. Metric: Mean benchmark score (%). Higher is better. Source: https://arxiv.org/html/2610.22356v1
Metric: Mean benchmark score (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DK-VLM-8B | 64.09 | |
| 2 | GeoThinker-8B | 62.23 | |
| 3 | DK-VLM-7B | 60.84 | |
| 4 | GeoThinker-7B | 60.43 | |
| 5 | VG-LLM-8B | 51.05 | |
| 6 | Qwen3-VL-8B (DK-VLM paper baseline) | 47.73 | |
| 7 | Cambrian-S-7B | 47.28 | |
| 8 | VLM-3R-7B | 45.4 | |
| 9 | Qwen2.5-VL-7B (DK-VLM paper baseline) | 40.31 |
Interactive version: theaggregate.ai/benchmark?slug=dk-vlm-six-benchmark-mean · How It Works · Data refreshed daily, snapshot 2026-10-17.