Research map / Vision-language and multimodal / Human and embodied modeling
Textual Grounding for Open-vocabulary Visual Information Extraction in Layout-diversified Documents
Mengjun Cheng, Chengquan Zhang, Chang Liu, Yuke Li, Bohan Li, Kun Yao, Xiawu Zheng, Rongrong Ji, Jie Chen
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 cross-modal semantic alignment pre-training stage (ablation) | TG-VIE · SVRD task3 test | H-mean: 39.52 -> 50.58 | Improved |
| add spatial embedding to the cross-region encoder (ablation) | TG-VIE · SVRD task3 test | H-mean: 46.58 -> 50.58 | Improved |
| add index (reading order) embedding (ablation) | TG-VIE · SVRD task3 test | H-mean: 48.95 -> 50.58 | Improved |
| add explicit segment extraction for layout modeling (ablation) | TG-VIE · SVRD task3 test | H-mean: 16.99 -> 39.52 | Improved |
| combine pixel-level and segment-level fusion (ablation) | TG-VIE · SVRD task3 test | H-mean: 37.12 -> 39.52 | Improved |
| label extra segment proposals as negative samples (ablation) | TG-VIE · SVRD task3 test | H-mean: 44.00 -> 50.58 | Improved |
| increase sampled query-segment pairs from 16 to 32 (ablation) | TG-VIE · SVRD task3 test | H-mean: 50.58 -> 48.79 | Hurt |
| use textual grounding instead of QA-style MLLM | TG-VIE vs QwenVL · SVRD task3 test | F-score: 17.63 -> 41.98 | Improved |
| use TG-VIE instead of tuned QwenVL on the paper document subset | TG-VIE · SVRD task3 test (paper subset) | F-score: 26.13 -> 7.74 | Hurt |
Closest papers by meaning
- Relation-Rich Visual Document Generator for Visual Information Extraction (CVPR 2025 · similarity 0.88)
- LDP: Generalizing to Multilingual Visual Information Extraction by Language Decoupled Pretraining (AAAI 2025 · similarity 0.87)
- A Simple yet Effective Layout Token in Large Language Models for Document Understanding (CVPR 2025 · similarity 0.87)
- SAIL: Sample-Centric In-Context Learning for Document Information Extraction (AAAI 2025 · similarity 0.87)
- Adaptive Markup Language Generation for Contextually-Grounded Visual Document Understanding (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)
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.
