Elamur Research⌕
☰
home / papers / 2026 / TableVista: Benchmarking Multimodal Table Reasoning under Visual and Structural Complexity

TableVista: Benchmarking Multimodal Table Reasoning under Visual and Structural Complexity

Paper · 2026 · Zheyuan Yang, Liqiang Shang, Junjie Chen et al.

Summary

TableVista asks how reliably multimodal foundation models reason over tables shown as images when both the layout and the visual presentation vary. Earlier benchmarks usually show each table in one clean rendering. The authors drew records from 14 public datasets, including WTQ, HiTab, TabFact, MMQA and FinQA. With 12 expert annotators and GPT-5 assistance, they relabeled, filtered and rewrote these records into 3,000 reasoning-heavy questions. The questions span five structural types: simple, text-mixed, complex/hierarchical, long and multi-table. A rendering pipeline then turns each question into 10 visual variants, giving 30,000 samples. The authors tested 29 open-source and proprietary models. Changing the rendering style barely affects results. Accuracy drops sharply, however, on complex structures, realistic captures such as photos, and fragmented layouts. Chain-of-thought prompting closes much of the gap between open and closed models, but large differences return under direct-output prompting. Error analysis shows that most failures come from spatial and structural misreading rather than OCR. TableVista therefore offers a robustness-focused tool for building structure-aware table understanding systems.

Key points

  • The 10 variants per question fall into three groups:
    • scenario styles: Web, LaTeX, Excel, Custom
    • perturbations: Noise, Structural, Partial, Missing
    • vision-only captures: Screenshot, Photo
  • Base Set curation involved several steps:
    • GPT-5 labeled each record for layout, information richness (1–5), four skill scores and reasoning-step count.
    • Annotators checked every label and corrected errors.
    • Questions were either rephrased or newly authored as multi-step questions.
  • Chain-of-thought prompting yields large gains over direct output:
    • Qwen3-VL-8B rises from 43.8 to 86.0.
    • GPT-5.4 rises from 72.1 to 95.6.
    • Under chain-of-thought, Qwen3.5-27B overtakes GPT-5.4 at the top of the leaderboard.
  • The breakdown of 100 sampled errors (from GPT-5.4 and Qwen3.5-27B) is:
    • Table Understanding: 54%, mostly Spatial Alignment (32%) and Structure Parsing (22%)
    • Reasoning & Calculation: 29%
    • Visual Perception: 12%
    • Pure text recognition errors: only 4%
  • Data and code are released as TableVista (repository FlowRays/TableVista). The work does not propose training methods to fix the failures it identifies.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Yale UniversityNew Haven · United States101 works · 13,302 citations🎓Tianjin UniversityTianjin · China52 works · 1,128 citations🎓Tongji UniversityShanghai · China33 works · 2,993 citations🎓University of BristolBristol · United Kingdom18 works · 822 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksComputer Vision and Pattern Recognition · 3,486 works

🧭 Topics 2Research subjects and languages the work is about.Reasoning · 1,685 worksVision & language · 811 works

Cite

Cite as Zheyuan Yang, Liqiang Shang, Junjie Chen, Xun Yang, Chenglong Xu, Bo Yuan, Chenyuan Jiao, Yaoru Sun and Yilun Zhao. 2026. TableVista: Benchmarking Multimodal Table Reasoning under Visual and Structural Complexity. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.1745

@inproceedings{yang2026tablevista,
  title = {{TableVista: Benchmarking Multimodal Table Reasoning under Visual and Structural Complexity}},
  author = {Zheyuan Yang, Liqiang Shang, Junjie Chen, Xun Yang, Chenglong Xu, Bo Yuan, Chenyuan Jiao, Yaoru Sun and Yilun Zhao},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1745},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1745}
}

References

37 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Iñigo Alonso, Imanol Miranda, Eneko Agirre, and Mirella Lapata. 2025. Tablet: A large-scale dataset for robust visual table understanding. ArXiv, abs/2509.21205. — [arXiv:2509.21205](https://arxiv.org/abs/2509.21205) find
  2. Shuai Bai, Yuxuan Cai, Ruizhe Chen, Keqin Chen, Xiong-Hui Chen, Zesen Cheng, Lianghao Deng, Wei Ding, Rongyao Fang, Chang Gao, Chunjiang Ge, Wenbin Ge, Zhifang Guo, Qidong Huang, Jie Huang, Fei Huang, Binyuan Hui, Shutong Jiang, Zhaohai Li, and 46 others. 2025a. Qwen3-vl technical report.
  3. Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, Humen Zhong, Yuanzhi Zhu, Mingkun Yang, Zhaohai Li, Jianqiang Wan, Pengfei Wang, Wei Ding, Zheren Fu, Yiheng Xu, and 8 others. 2025b. Qwen2.5-vl technical report. ArXiv, abs/2502.13923.
  4. Wenhu Chen, Hongmin Wang, Jianshu Chen, Yunkai Zhang, Hong Wang, SHIYANG LI, Xiyou Zhou, and William Yang Wang. 2019. Tabfact: A large-scale dataset for table-based fact verification. ArXiv, abs/1909.02164.
  5. Zhiyu Chen, Wenhu Chen, Charese Smiley, Sameena Shah, Iana Borova, Dylan Langdon, Reema Moussa, Matthew I. Beane, Ting-Hao ’Kenneth’ Huang, Bryan R. Routledge, and William Yang Wang. 2021. Finqa: A dataset of numerical reasoning over financial data. ArXiv, abs/2109.00122.
  6. Zhoujun Cheng, Haoyu Dong, Zhiruo Wang, Ran Jia, Jiaqi Guo, Yan Gao, Shi Han, Jian-Guang Lou, and Dongmei Zhang. 2021. Hitab: A hierarchical table dataset for question answering and natural language generation. In Annual Meeting of the Association for Computational Linguistics.
  7. Google DeepMind. 2026. Gemma 4. find
  8. DeepSeek-AI, Daya Guo, Dejian Yang, Haowei Zhang, Jun-Mei Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiaoling Bi, Xiaokang Zhang, Xingkai Yu, Yu Wu, Z. F. Wu, Zhibin Gou, Zhihong Shao, Zhuoshu Li, Ziyi Gao, and 179 others. 2025. Deepseek-r1: Incentivizing reasoning capability i…
  9. Naihao Deng, Zhenjie Sun, Ruiqi He, Aman Sikka, Yulong Chen, Lin Ma, Yue Zhang, and Rada Mihalcea. 2024. Tables as texts or images: Evaluating the table reasoning ability of llms and mllms. In Annual Meeting of the Association for Computational Linguistics. find
  10. Allen Institute for AI. 2025. Molmo2: Open weights and data for vision-language models with video understanding and grounding. find
  11. Jun-Peng Jiang, Yu Xia, Hai-Long Sun, Shiyin Lu, Qing-Guo Chen, Weihua Luo, Kaifu Zhang, De-Chuan Zhan, and Han-Jia Ye. 2025. Multimodal tabular reasoning with privileged structured information. ArXiv, abs/2506.04088. — [arXiv:2506.04088](https://arxiv.org/abs/2506.04088) find
  12. Gemma Team Aishwarya Kamath, Johan Ferret, Shreya Pathak, Nino Vieillard, Ramona Merhej, Sarah Perrin, Tatiana Matejovicova, Alexandre Ram’e, Morgane Rivière, Louis Rouillard, Thomas Mesnard, Geoffrey Cideron, Jean-Bastien Grill, Sabela Ramos, Edouard Yvinec, Michelle Casbon, Etienne Pot, Ivo Penche… find
  13. Yoonsik Kim, Moonbin Yim, and Ka Yeon Song. 2024. Tablevqa-bench: A visual question answering benchmark on multiple table domains. ArXiv, abs/2404.19205. — [arXiv:2404.19205](https://arxiv.org/abs/2404.19205) find
  14. Haotian Liu, Chunyuan Li, Yuheng Li, and Yong Jae Lee. 2023. Improved baselines with visual instruction tuning.
  15. Haotian Liu, Chunyuan Li, Yuheng Li, Bo Li, Yuanhan Zhang, Sheng Shen, and Yong Jae Lee. 2024. Llava-next: Improved reasoning, ocr, and world knowledge. find
  16. Suyash Vardhan Mathur, Jainit Sushil Bafna, Kunal Kartik, Harshita Khandelwal, Manish Shrivastava, Vivek Gupta, Mohit Bansal, and Dan Roth. 2024. Knowledge-aware reasoning over multimodal semi-structured tables. In Conference on Empirical Methods in Natural Language Processing. find
  17. Linyong Nan, Chia-Hsuan Hsieh, Ziming Mao, Xi Victoria Lin, Neha Verma, Rui Zhang, Wojciech Kryscinski, Nick Schoelkopf, Riley Kong, Xiangru Tang, Murori Mutuma, Benjamin Rosand, Isabel Trindade, Renusree Bandaru, Jacob Cunningham, Caiming Xiong, and Dragomir R. Radev. 2021. Fetaqa: Free-form table…
  18. Panupong Pasupat and Percy Liang. 2015. Compositional semantic parsing on semi-structured tables. In Annual Meeting of the Association for Computational Linguistics.
  19. Qwen Team. 2026. Qwen3.6-35B-A3B: Agentic coding power, now open to all. find
  20. Qwen Team. 2026. Qwen3.5: Accelerating productivity with native multimodal agents. find
  21. Prasham Titiya, Jainil Trivedi, Chitta Baral, and Vivek Gupta. 2025. Mmtbench: A unified benchmark for complex multimodal table reasoning. ArXiv, abs/2505.21771. — [arXiv:2505.21771](https://arxiv.org/abs/2505.21771) find
  22. Lanrui Wang, Mingyu Zheng, Hongyin Tang, Zheng Lin, Yanan Cao, Jingang Wang, Xunliang Cai, and Weiping Wang. 2025a. Needleinatable: Exploring long-context capability of large language models towards long-structured tables. ArXiv, abs/2504.06560. — [arXiv:2504.06560](https://arxiv.org/abs/2504.06560) find
  23. Weiyun Wang, Zhangwei Gao, Lixin Gu, Hengjun Pu, Long Cui, Xingguang Wei, Zhaoyang Liu, Linglin Jing, Shenglong Ye, Jie Shao, Zhaokai Wang, Zhe Chen, Hongjie Zhang, Ganlin Yang, Haomin Wang, Qi Wei, Jinhui Yin, Wenhao Li, Erfei Cui, and 44 others. 2025b. Internvl3.5: Advancing open-source multimodal…
  24. Jian Wu, Linyi Yang, Dongyuan Li, Yuliang Ji, Manabu Okumura, and Yue Zhang. 2025. Mmqa: Evaluating llms with multi-table multi-hop complex questions. In International Conference on Learning Representations. find
  25. Bohao Yang, Yingji Zhang, Dong Liu, Andr’e Freitas, and Chenghua Lin. 2025a. Does table source matter? benchmarking and improving multimodal scientific table understanding and reasoning. ArXiv, abs/2501.13042. — [arXiv:2501.13042](https://arxiv.org/abs/2501.13042) find
  26. Zheyuan Yang, Lyuhao Chen, Arman Cohan, and Yilun Zhao. 2025b. Table-r1: Inference-time scaling for table reasoning. ArXiv, abs/2505.23621. — [arXiv:2505.23621](https://arxiv.org/abs/2505.23621) find
  27. Tianyu Yu, Zefan Wang, Chongyi Wang, Fuwei Huang, Wenshuo Ma, Zhihui He, Tianchi Cai, Weize Chen, Yuxiang Huang, Yuanqian Zhao, Bokai Xu, Junbo Cui, Ying Xu, Liqing Ruan, Lu Zhang, Hanyu Liu, Jing Tang, Hongyuan Liu, Qi Guo, and 15 others. 2025. Minicpm-v 4.5: Cooking efficient mllms via architectur…
  28. Yilun Zhao, Zhenting Qi, Linyong Nan, Boyu Mi, Yixin Liu, Weijin Zou, Simeng Han, Ruizhe Chen, Xiangru Tang, Yumo Xu, Dragomir Radev, and Arman Cohan. 2023a. QTSumm: Query-focused summarization over tabular data. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processi… find
  29. Yilun Zhao, Chen Zhao, Linyong Nan, Zhenting Qi, Wenlin Zhang, Xiangru Tang, Boyu Mi, and Dragomir Radev. 2023b. RobuT: A systematic study of table QA robustness against human-annotated adversarial perturbations. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguist… find
  30. Mingyu Zheng, Xinwei Feng, Qingyi Si, Qiaoqiao She, Zheng Lin, Wenbin Jiang, and Weiping Wang. 2024. Multimodal table understanding. In Annual Meeting of the Association for Computational Linguistics.
  31. Bangbang Zhou, Zuan Gao, Zixiao Wang, Boqiang Zhang, Yuxin Wang, Zhineng Chen, and Hongtao Xie. 2025a. Syntab-llava: Enhancing multimodal table understanding with decoupled synthesis. 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 24796–24806. find
  32. Wei Zhou, Mohsen Mesgar, Heike Adel, and Annemarie Friedrich. 2025b. Texts or images? a fine-grained analysis on the effectiveness of input representations and models for table question answering. ArXiv, abs/2505.14131. find
  33. Junwen Zhang, Pu Chen, and Yin Zhang. 2025. Table-moe: Neuro-symbolic routing for structured expert reasoning in multimodal table understanding. ArXiv, abs/2506.21393. — [arXiv:2506.21393](https://arxiv.org/abs/2506.21393) find
  34. Tianshu Zhang, Xiang Yue, Yifei Li, and Huan Sun. 2023. Tablellama: Towards open large generalist models for tables. ArXiv, abs/2311.09206.
  35. Weichao Zhao, Hao Feng, Qi Liu, Jingqun Tang, Shubo Wei, Binghong Wu, Lei Liao, Yongjie Ye, Hao Liu, Houqiang Li, and Can Huang. 2024a. Tabpedia: Towards comprehensive visual table understanding with concept synergy. ArXiv, abs/2406.01326. — [arXiv:2406.01326](https://arxiv.org/abs/2406.01326) find
  36. Yilun Zhao, Yitao Long, Hongjun Liu, Ryo Kamoi, Linyong Nan, Lyuhao Chen, Yixin Liu, Xiangru Tang, Rui Zhang, and Arman Cohan. 2024b. DocMath-eval: Evaluating math reasoning capabilities of LLMs in understanding long and specialized documents. In Proceedings of the 62nd Annual Meeting of the Associa… find
  37. ID Education Major Assigned Task PhD Student (Yr 3) Computer Science Master (Yr 2) Computer Science Attribute Labeling, QA Enhancement, Visual Rendering Audit, Quality Audit Attribute Labeling, QA Enhancement PhD Student (Yr 2) Computer Science Attribute Labeling, QA Enhancement, Visual Rendering Au… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yang-2026-tablevista-benchmarking-multimodal-table-reasoning/