Research map / Large language models / Language, safety and interpretability
SmolDocling: An ultra-compact vision-language model for end-to-end multi-modal document conversion
Ahmed Nassar, Matteo Omenetti, Maksym Lysak, Nikolaos Livathinos, Christoph Auer, Lucas Morin, Rafael Teixeira de Lima, Yusik Kim, A. Said Gurbuz, Michele Dolfi, Peter W. J. Staar
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 |
|---|---|---|---|
| use 256M SmolDocling with DocTags instead of larger VLMs for full-page conversion | SmolDocling · DocLayNet full-page text | Edit Distance: 0.56 -> 0.48 vs Qwen2.5 VL 7B | Improved |
| SmolDocling vs Qwen2.5-VL on layout analysis | SmolDocling · DocLayNet 6-class | mAP[0.5:0.95]: 0.133 -> 0.231 | Improved |
| SmolDocling vs Qwen2.5-VL on Picture class layout | SmolDocling · DocLayNet Picture class | mAP[0.5:0.95]: 0.078 -> 0.066 | Hurt |
| SmolDocling vs specialised TableFormer on table structure with text | SmolDocling · FinTabNet | TEDS: 0.89 -> 0.52 | Hurt |
| SmolDocling vs Granite Vision on chart-to-table | SmolDocling · Chart-to-table | TEDS: 0.95 -> 0.75 | Hurt |
| SmolDocling vs GOT on equation recognition | SmolDocling · Im2Latex-230k | Edit Distance: 0.11 vs 0.11 | No clear effect |
| SmolDocling vs Qwen2.5 VL on equation recognition | SmolDocling · Im2Latex-230k | Edit Distance: 0.22 -> 0.11 | Improved |
Closest papers by meaning
- Towards Real-World Document Parsing via Realistic Scene Synthesis and Document-Aware Training (CVPR 2026 · similarity 0.86)
- MosaicDoc: A Large-Scale Bilingual Benchmark for Visually Rich Document Understanding (AAAI 2026 · similarity 0.85)
- DocLayLLM: An Efficient Multi-modal Extension of Large Language Models for Text-rich Document Understanding (CVPR 2025 · similarity 0.85)
- Adaptive Markup Language Generation for Contextually-Grounded Visual Document Understanding (CVPR 2025 · similarity 0.85)
- A Simple yet Effective Layout Token in Large Language Models for Document Understanding (CVPR 2025 · similarity 0.85)
- Boosting Document Parsing Efficiency and Performance with Coarse-to-Fine Visual Processing (CVPR 2026 · similarity 0.85)
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.
