Research map / Vision-language and multimodal / Multimodal LLMs
StrucTab: A Structured Optimization Framework for Table Parsing
Gengluo Li, Shangpin Peng, Chengquan Zhang, Binghong Wu, Hao Feng, Weinong Wang, Pengyuan Lyu, Huawen Shen, Xingyu Wan, Zhuotao Tian, Han Hu, Can Ma and 1 more
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 row-column counting subtask (ablation) | StrucTab baseline · Open-source benchmarks average (OmniDocBench, CC-OCR, OCRBench v2) | TEDS: 86.26 -> 87.43 | Improved |
| add merged-cell analysis subtask (ablation) | StrucTab baseline · Open-source benchmarks average (OmniDocBench, CC-OCR, OCRBench v2) | TEDS: 86.26 -> 87.39 | Improved |
| combine row-column counting and merged-cell analysis (ablation) | StrucTab baseline · Open-source benchmarks average (OmniDocBench, CC-OCR, OCRBench v2) | TEDS: 87.43 -> 88.93 | Improved |
| unify subtasks into sequential reasoning (ablation) | independent subtasks (d) · Open-source benchmarks average (OmniDocBench, CC-OCR, OCRBench v2) | TEDS: 88.93 -> 90.11 | Improved |
| add first-stage curriculum of independent subtasks before sequential reasoning (ablation) | Seq reasoning · Open-source benchmarks average (OmniDocBench, CC-OCR, OCRBench v2) | TEDS: 89.54 -> 90.11 | Improved |
| add Uni-TabRL to SFT model | StrucTab-SFT · Average TEDS across four benchmarks | TEDS: 85.87 -> 88.79 | Improved |
| add GRPO RL with TEDS rewards (ablation) | StrucTab-SFT · Open-source benchmarks average (OmniDocBench, CC-OCR, OCRBench v2) | TEDS: 90.11 -> 91.29 | Improved |
| replace TEDS content reward with VLM consistency judge (ablation) | RL baseline · Open-source benchmarks average (OmniDocBench, CC-OCR, OCRBench v2) | TEDS: 91.29 -> 90.51 | Hurt |
| replace TEDS content reward with Anchor-Guided Destylization (ablation) | RL baseline · Open-source benchmarks average (OmniDocBench, CC-OCR, OCRBench v2) | TEDS: 91.29 -> 91.76 | Improved |
| replace TEDS-S structure reward with 1D Probe (ablation) | RL baseline · Open-source benchmarks average (OmniDocBench, CC-OCR, OCRBench v2) | TEDS: 91.29 -> 91.80 | Improved |
| combine Anchor-Guided Destylization and 1D Probe (ablation) | RL baseline · TableVerse-5K | TEDS: 86.45 -> 87.81 | Improved |
| use 1D Probe structure reward instead of TEDS-S (ablation) | RL · TableVerse-5K | repetition rate: 1.20% -> 1.02% | Improved |
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.
- TRivia-3B — TRivia: Self-supervised Fine-tuning of Vision-Language Models for Table Recognition (CVPR 2026) · better on 1 of 1 shared result
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
- Decoupling Skeleton and Flesh: Efficient Multimodal Table Reasoning with Disentangled Alignment and Structure-aware Guidance (ICML 2026 · similarity 0.88)
- NeedleInATable: Exploring Long-Context Capability of Large Language Models towards Long-Structured Tables (NeurIPS 2025 · similarity 0.87)
- Multimodal Tabular Reasoning with Privileged Structured Information (NeurIPS 2025 · similarity 0.87)
- TableDART: Dynamic Adaptive Multi-Modal Routing for Table Understanding (ICLR 2026 · similarity 0.87)
- STRABLE: Benchmarking Tabular Machine Learning with Strings (NeurIPS 2026 · similarity 0.86)
- TRivia: Self-supervised Fine-tuning of Vision-Language Models for Table Recognition (CVPR 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.
