DRAGON (Diagrams) - MapWise: leaderboard
DRAGON (2026) evidence-grounded reasoning on maps from MapWise: given the question and its answer, the model must box every diagram region needed to justify the answer. Grounding IoU (%) between predicted and human-verified evidence regions, averaged over three prompting strategies; higher is better.
Metric: Evidence grounding on maps from MapWise: given the diagram, the question and the verified answer, the model returns bounding boxes for every visual element needed to justify the answer; Grounding IoU (%) between the union of predicted boxes and the union of the human-verified gold evidence boxes, averaged over the EDGE (direct), SAGE (select then ground) and VERGE (verify and refine) prompting strategies; DRAGON test split; temperature 0.2, top-p 0.7; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 8 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2.5 | 10.8 | #272 |
| 2 | Claude Sonnet 4.6 | 9.7 | #179 |
| 3 | Gemini 3 Pro (Preview) | 8.5 | #123 |
| 4 | Claude Opus 4.6 | 8.3 | #114 |
| 5 | Llama 4 Maverick Instruct | 5.2 | #1126 |
| 6 | Gemma 3 27B (IT) | 5.1 | #1341 |
| 7 | Qwen 3.5 35B A3B | 4.3 | #494 |
| 8 | InternVL3.5-38B | 2.7 | #1026 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=dragon-diagrams-mapwise · How It Works · Data refreshed daily, snapshot 2026-11-22.