Research map / Vision-language and multimodal / Multimodal LLMs
GroundSet: A Cadastral-Grounded Dataset for Spatial Understanding with Vector Data
Roger Ferrod, Maël Lecene, Krishna Sapkota, George Leifman, Vered Silverman, Genady Beryozkin, Sylvain Lobry
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 |
|---|---|---|---|
| fine-tune LLaVA-1.6 on GroundSet instructions | LLaVA-1.6 7B · GroundSet test, Classification | acc@0.8: 29.20 -> 94.18 | Improved |
| fine-tune LLaVA-1.6 on GroundSet instructions | LLaVA-1.6 7B · GroundSet test, Detection | F1@0.5: 0.92 -> 49.47 | Improved |
| fine-tune LLaVA-1.6 on GroundSet instructions | LLaVA-1.6 7B · GroundSet test, Captioning | CIDEr: 17.93 -> 104.48 | Improved |
| RS-specific fine-tuning on legacy data (GeoChat) vs general model | LLaVA-1.5 base · GroundSet test, Classification | acc@0.8: 28.62 -> 7.59 | Hurt |
| RS-specific fine-tuning (SkySenseGPT) vs general LLaVA on VQA | LLaVA base · GroundSet test, VQA | F1: 86.36 -> 89.69 | Mixed |
| use grounding-specialized architecture (Ferret) vs RS specialists | Ferret · GroundSet test, Detection | F1@0.5: 5.52 -> 18.08 | Improved |
| add text-only polygon reasoning augmentation (ablation) | GroundSet LLaVA-1.6 · 100k subset, Segmentation | F1@0.5: 12.62 -> 14.52 | Improved |
| add exhaustive segmentation queries (ablation) | GroundSet LLaVA-1.6 · 100k subset, Segmentation | F1@0.5: 12.62 -> 14.54 | Improved |
| add text-only geometric augmentation (ablation) | GroundSet LLaVA-1.6 · 100k subset, Classification | acc@0.8: 90.47 -> 90.06 | No clear effect |
| unfreeze vision encoder (ablation) | GroundSet LLaVA-1.6 · 100k subset, Segmentation | F1@0.5: 12.62 -> 11.52 | Hurt |
| filter occluded invisible instances (ablation) | GroundSet LLaVA-1.6 · 100k subset, Segmentation | F1@0.5: 12.62 -> 15.31 | Mixed |
| train on GroundSet, evaluate zero-shot on VRSBench detection | GroundSet LLaVA-1.6 · VRSBench, Detection | F1@0.5: 12.81 -> 15.47 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Ferret — Ferret: Refer and Ground Anything Anywhere at Any Granularity (ICLR 2024) · better on 3 of 3 shared results
- GeoChat — GeoChat: Grounded Large Vision-Language Model for Remote Sensing (CVPR 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- GeoViS: Geospatially Rewarded Visual Search for Remote Sensing Visual Grounding (CVPR 2026 · similarity 0.89)
- TerraScope: Pixel-Grounded Visual Reasoning for Earth Observation (CVPR 2026 · similarity 0.87)
- GeoPixel: Pixel Grounding Large Multimodal Model in Remote Sensing (ICML 2025 · similarity 0.87)
- GeoCore-9B: Towards Geo-Aware Generative Foundation Models in Earth Observation (NeurIPS 2026 · similarity 0.86)
- Towards Faithful Reasoning in Remote Sensing: A Perceptually-Grounded GeoSpatial Chain-of-Thought for Vision-Language Models (ICLR 2026 · similarity 0.86)
- Reasoning in Space via Grounding in the World (ICLR 2026 · similarity 0.86)
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.
