Research map / Vision-language and multimodal / Multimodal LLMs
DOGR: Towards Versatile Visual Document Grounding and Referring
Yinan Zhou, Yuxin Chen, Haokun Lin, Yichen Wu, Shuyu Yang, Zhongang Qi, Chen Ma, Li Zhu
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 multi-granular parsing data in pre-training (ablation) | DOGR · DocLocal4K word localization | IoU@0.5 accuracy: 74.79 -> 81.85 | Improved |
| add multi-granular parsing data in pre-training (ablation) | DOGR · DocLocal4K line recognition | BLEU-4: 73.88 -> 77.60 | Improved |
| add multi-granular parsing data in pre-training (ablation) | DOGR · DocLocal4K paragraph localization | IoU@0.5 accuracy: 95.46 -> 95.67 | No clear effect |
| add multi-granular parsing data in pre-training (ablation) | DOGR · DOGR-Local15k chart grounding | accuracy: 24.03 -> 86.12 | Improved |
| add multi-granular parsing data in pre-training (ablation) | DOGR · DOGR-Local15k poster grounding | accuracy: 63.19 -> 93.53 | Improved |
| add DOGR instruction-tuning data to DocOwl-1.5 data (ablation) | DOGR · InfoVQA | ANLS: 64.58 -> 67.45 | Improved |
| add DOGR instruction-tuning data to DocOwl-1.5 data (ablation) | DOGR · VisualMRC | CIDEr: 265.9 -> 287.25 | Improved |
| add DOGR instruction-tuning data to DocOwl-1.5 data (ablation) | DOGR · DocVQA | ANLS: 87.57 -> 89.24 | Improved |
| use accurate re-rendered bboxes instead of perturbed bboxes in chart data (ablation) | DOGR · DOGR-Bench chart Gr | F1_all: 20.2 -> 32.91 | Improved |
| generate chart QA with GPT-4o instead of Gemini 1.5 flash (ablation) | DOGR · DOGR-Bench chart Gr | F1_all: 1.10 -> 32.91 | Improved |
| add 25k GPT-4o chart data with perturbed bboxes (ablation) | DOGR · ChartQA | accuracy: 80.92 -> 81.42 | Improved |
| add 25k GPT-4o chart data with accurate bboxes (ablation) | DOGR · ChartQA | accuracy: 80.92 -> 81.88 | Improved |
Closest papers by meaning
- Grounding Multimodal Large Language Models to the World (ICLR 2024 · similarity 0.88)
- DocSeeker: Structured Visual Reasoning with Evidence Grounding for Long Document Understanding (CVPR 2026 · similarity 0.87)
- GroundingME: Exposing the Visual Grounding Gap in MLLMs through Multi-Dimensional Evaluation (CVPR 2026 · similarity 0.87)
- Docopilot: Improving Multimodal Models for Document-Level Understanding (CVPR 2025 · similarity 0.86)
- DocVAL: Validated Chain-of-Thought Distillation for Grounded Document VQA (ICML 2026 · similarity 0.86)
- DocHop: Benchmarking Out-of-domain Multi-hop Reasoning in Information-Dense Documents (ICML 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.
