Research map / Vision-language and multimodal / Human and embodied modeling
M3DocDep: Multi-modal, Multi-page, Multi-document Dependency Chunking with Large Vision-Language Models
Joongmin Shin, Jeongbae Park, Jaehyung Seo, Heuiseok Lim
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 |
|---|---|---|---|
| add MST global tree decoding instead of local argmax (ablation) | M3DocDep · DHP macro-average | STEDS: 66.30 -> 73.00 | Improved |
| allow cross-page parent edges (ablation) | M3DocDep · DHP macro-average | STEDS: 63.74 -> 73.00 | Improved |
| recover dependency tree before chunking vs MultiDocFusion | M3DocDep · MOAMOB retrieval | nDCG: 65.54 -> 75.54 | Improved |
| recover dependency tree before chunking vs MultiDocFusion | M3DocDep · CUAD retrieval | nDCG: 88.19 -> 89.12 | Improved |
| tree-guided chunking vs MultiDocFusion | M3DocDep · DUDE RAG QA | ANLS: 18.59 -> 21.43 | Improved |
| remove section-path and page-range metadata from both methods (ablation) | M3DocDep vs MultiDocFusion · corpus retrieval | nDCG | Improved |
| use biaffine edge scoring on LVLM block embeddings vs SFT generation | M3DocDep · DocHieNet | STEDS: 40.39 -> 70.83 | Improved |
| swap LVLM encoder (Qwen2.5-VL, InternVL-3.5, LLaVA-OV-1.5) (ablation) | M3DocDep · DocHieNet | F1: 76.01 / 75.71 / 74.07 | No clear effect |
Closest papers by meaning
- Towards Real-World Document Parsing via Realistic Scene Synthesis and Document-Aware Training (CVPR 2026 · similarity 0.88)
- MPDocBench-Parse: Benchmarking Practical Multi-page Document Parsing (NeurIPS 2026 · similarity 0.87)
- VDocRAG: Retrieval-Augmented Generation over Visually-Rich Documents (CVPR 2025 · similarity 0.86)
- DocLayLLM: An Efficient Multi-modal Extension of Large Language Models for Text-rich Document Understanding (CVPR 2025 · similarity 0.86)
- SV-RAG: LoRA-Contextualizing Adaptation of MLLMs for Long Document Understanding (ICLR 2025 · similarity 0.86)
- VisRAG: Vision-based Retrieval-augmented Generation on Multi-modality Documents (ICLR 2025 · 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.
