Research map / Vision-language and multimodal / Multimodal LLMs
OmniParser V2: Structured-Points-of-Thought for Unified Visual Text Parsing and Its Generality to Multimodal Large Language Models
Wenwen Yu, Zhibo Yang, Jianqiang Wan, Sibo Song, Jun Tang, Wenqing Cheng, Yuliang Liu, Xiang Bai
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 spatial-window prompting in pre-training (ablation) | OmniParser V2 · Total-Text | E2E None F1: 82.5 -> 83.8 | Improved |
| add prefix-window prompting in pre-training (ablation) | OmniParser V2 · Total-Text | E2E None F1: 82.5 -> 83.2 | Improved |
| add both spatial- and prefix-window prompting (ablation) | OmniParser V2 · Total-Text | E2E None F1: 82.5 -> 84.3 | Improved |
| replace naive shared decoder with token-router decoder (ablation) | OmniParser V2 · Total-Text / ICDAR 2015 | text spotting average: +1.72% over naive shared decoder | Improved |
| use token-router decoder instead of OmniParser's non-shared decoder (ablation) | OmniParser V2 · Total-Text / ICDAR 2015 | FPS and params: 2.6 -> 8.2 FPS, 144 -> 110 params (M) | Improved |
| increase C-decoder length from 200 to 300 and S-decoder length from 1,500 to 2,000 (ablation) | OmniParser V2 · PubTabNet | TEDS: 88.9 -> 89.0 | No clear effect |
| shorten S-decoder length from 1,500 to 1,124 (ablation) | OmniParser V2 · PubTabNet | TEDS: 88.9 -> 88.2 | Hurt |
| add layout analysis SPOT to native text spotting SPOT (ablation) | OmniParser V2 · SignaTR6K | mean F1: 69.8 -> 82.1 | Improved |
| add layout analysis SPOT (handwritten text) (ablation) | OmniParser V2 · SignaTR6K | HT F1: 61.5 -> 79.3 | Improved |
| add SPOT prompting to Qwen2.5-VL-3B | Qwen2.5-VL-3B · HallusionBench | accuracy: 63.8 -> 68.5 | Improved |
| add SPOT prompting to InternVL2.5-2B | InternVL2.5-2B · CORD | F1: 74.5 -> 83.7 | Improved |
| add SPOT prompting to InternVL2.5-2B | InternVL2.5-2B · SROIE | F1: 71.9 -> 77.8 | Improved |
Later papers that compared with this method
1 later accepted paper reports OmniParser V2 as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- ScreenParse: Moving Beyond Sparse Grounding with Complete Screen Parsing Supervision (ICML 2026 · better on 3 of 3 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- OmniParser (V1) — OmniParser: A Unified Framework for Text Spotting Key Information Extraction and Table Recognition (CVPR 2024) · better on 10 of 11 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
- OmniParser: A Unified Framework for Text Spotting Key Information Extraction and Table Recognition (CVPR 2024 · similarity 0.95)
- Read, Parse, Describe: Unified Document Parsing with Visual Element Description Generation (NeurIPS 2026 · similarity 0.88)
- Towards Real-World Document Parsing via Realistic Scene Synthesis and Document-Aware Training (CVPR 2026 · similarity 0.87)
- Boosting Document Parsing Efficiency and Performance with Coarse-to-Fine Visual Processing (CVPR 2026 · similarity 0.86)
- Magnifying What Matters: Attention-Guided Adaptive Rendering for Visual Text Comprehension (NeurIPS 2026 · similarity 0.86)
- Marten: Visual Question Answering with Mask Generation for Multi-modal Document Understanding (CVPR 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.
