Research map / Image generation / Autoregressive and latent generation
MARS - A Foundational Map Auto-Regressor
Qi Zhang, Suvam Bag, Rupanjali Kukal, Mikael Figueroa, Rishi Madhok, Nikolaos Karianakis, Fuxun Yu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| pretrain on MAP-3M before fine-tuning (ablation) | MARS · SpaceNet roads | F1: 70.45 -> 82.05 | Improved |
| pretrain on MAP-3M before fine-tuning (ablation) | MARS · AICrowd buildings | IoU: 95.09 -> 95.24 | No clear effect |
| train on both road and building classes instead of road only (ablation) | MARS · MAP3M-VAL roads | F1: 87.7 -> 83.1 | Hurt |
| train on both classes instead of road only (ablation) | MARS · Cityscale roads | F1: 81.5 -> 82.9 | Improved |
| train on both classes instead of building only (ablation) | MARS · AICrowd buildings | IoU: 95.0 -> 97.3 | Improved |
| train on both classes instead of single class (ablation) | MARS · SpaceNet roads | F1: 81.2 -> 82.1 | Improved |
| add one-click chat prompt (Chat 1pt) | MARS · Cityscale roads | F1: 82.88 -> 83.79 | Improved |
| add two-click chat prompt (Chat 2pts) | MARS · Cityscale roads | F1: 82.88 -> 84.13 | Improved |
| add two-click chat prompt (Chat 2pts) | MARS · AICrowd buildings (partial) | IoU: 97.32 -> 97.40 | No clear effect |
| autoregressive MARS vs segmentation-based SamRoad | MARS · Cityscale roads | recall: 67.69 -> 81.53 | Improved |
| MARS vs RNGDet++ on SpaceNet | MARS · SpaceNet roads | F1: 82.51 -> 82.05 | No clear effect |
Closest papers by meaning
- Mars-Bench: A Benchmark for Evaluating Foundation Models for Mars Science Tasks (NeurIPS 2025 · similarity 0.85)
- MARS: Mixture of Auto-Regressive Models for Fine-grained Text-to-image Synthesis (AAAI 2025 · similarity 0.84)
- SDTagNet: Leveraging Text-Annotated Navigation Maps for Online HD Map Construction (NeurIPS 2025 · similarity 0.84)
- Vision-Language Reasoning for Geolocalization: A Reinforcement Learning Approach (AAAI 2026 · similarity 0.84)
- Socratic-Geo: Synthetic Data Generation and Cross-Modal Geometric Reasoning via Multi-Agent Interaction (CVPR 2026 · similarity 0.84)
- MapEval: A Map-Based Evaluation of Geo-Spatial Reasoning in Foundation Models (ICML 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
