Elamur Research⌕
☰
home / papers / 2026 / UnAC: Adaptive Visual Prompting with Abstraction and Stepwise Checking for Complex Multimodal Reasoning

UnAC: Adaptive Visual Prompting with Abstraction and Stepwise Checking for Complex Multimodal Reasoning

Paper · 2026 · Yifan Wang and Yun Fu

Summary

Large multimodal models (LMMs) perceive images well but often fail at multi-step reasoning grounded in visual evidence. The authors attribute many errors to misreading images or summarizing them poorly. They propose UnAC (Understanding, Abstracting, and Checking), a training-free prompting framework with three stages. First, the model analyzes the question to decide what it needs from the image. That need is sorted into semantic objects, which are marked with segmentation (SEEM), or textual symbols, which are marked with OCR, and the marks are overlaid as adaptive visual prompts. Second, the model turns the image into question-focused text, describing it globally and then examining the relevant marked regions. Third, a gradual self-checking step splits the problem into sub-questions and verifies each answer against the image, rather than reviewing the whole solution at once. UnAC improves results on MathVista, MM-Vet and MMMU for both large and lightweight LMMs, suggesting it is model-agnostic. The authors find that the LMM's own reasoning ability remains the main bottleneck.

Key points

  • Evaluated on MathVista, MM-Vet and MMMU with GPT-4V, Gemini-1.5-Flash and LLaVA-v1.6-7B/13B. Gains were reported for every model–dataset pair.
  • The largest highlighted gain is +6.4% absolute accuracy on MathVista with Gemini-1.5-Flash.
  • On the MathVista testmini subset with GPT-4V, overall accuracy was:
    • 52.7 without checking,
    • 53.4 with global checking,
    • 57.6 with gradual checking.
  • Using segmentation-only or OCR-only prompts helped little. Error analysis of global checking found that misdirection (50%) and math errors (39%) dominate.
  • Swapping in GPT-4V for LLaVA-v1.6-7B in the abstraction and checking stages raised accuracy, but the final conclusion step still limits results. The authors also note that visual prompts help little on highly abstract tasks such as geometry problems.

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

🎓Northeastern UniversityBoston · United States57 works · 6,533 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksComputer Vision and Pattern Recognition · 2,869 works

🧭 Topics 3Research subjects and languages the work is about.Language models & pretraining · 3,320 worksReasoning · 1,426 worksVision & language · 657 works

Cite

Cite as Yifan Wang and Yun Fu. 2026. UnAC: Adaptive Visual Prompting with Abstraction and Stepwise Checking for Complex Multimodal Reasoning. 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.1196

@inproceedings{wang2026unac,
  title = {{UnAC: Adaptive Visual Prompting with Abstraction and Stepwise Checking for Complex Multimodal Reasoning}},
  author = {Yifan Wang and Yun Fu},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1196},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1196}
}

References

35 references, 24 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, and 1 others. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774.
  2. Rohan Anil, Andrew M Dai, Orhan Firat, Melvin Johnson, Dmitry Lepikhin, Alexandre Passos, Siamak Shakeri, Emanuel Taropa, Paige Bailey, Zhifeng Chen, and 1 others. 2023. Palm 2 technical report. arXiv preprint arXiv:2305.10403.
  3. Bo Li, Yuanhan Zhang, Dong Guo, Renrui Zhang, Feng Li, Hao Zhang, Kaichen Zhang, Peiyuan Zhang, Yanwei Li, Ziwei Liu, and 1 others. 2024. Llava-onevision: Easy visual task transfer. arXiv preprint arXiv:2408.03326.
  4. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, and 1 others. 2020. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901.
  5. Zhan Ling, Yunhao Fang, Xuanlin Li, Zhiao Huang, Mingu Lee, Roland Memisevic, and Hao Su. 2024. Deductive verification of chain-of-thought reasoning. Advances in Neural Information Processing Systems, 36.
  6. Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, and 1 others. 2023. Sparks of artificial general intelligence: Early experiments with gpt-4. arXiv preprint arXiv:2303.12712.
  7. Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. 2024. Visual instruction tuning. Advances in neural information processing systems, 36.
  8. Xi Chen, Zhiyan Zhao, Yilei Zhang, Manni Duan, Donglian Qi, and Hengshuang Zhao. 2022. Fo-calclick: Towards practical interactive image segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1300–1309. find
  9. Zhe Chen, Jiannan Wu, Wenhai Wang, Weijie Su, Guo Chen, Sen Xing, Muyan Zhong, Qinglong Zhang, Xizhou Zhu, Lewei Lu, and 1 others. 2024. Internvl: Scaling up vision foundation models and aligning for generic visual-linguistic tasks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pa…
  10. Aakanksha Chowdhery, Sharan Narang, Jacob Devlin, Maarten Bosma, Gaurav Mishra, Adam Roberts, Paul Barham, Hyung Won Chung, Charles Sutton, Sebastian Gehrmann, and 1 others. 2023. Palm: Scaling language modeling with pathways. Journal of Machine Learning Research, 24(240):1–113.
  11. Wenliang Dai, Junnan Li, Dongxu Li, Anthony Meng Huat Tiong, Junqi Zhao, Weisheng Wang, Boyang Li, Pascale N Fung, and Steven Hoi. 2024. Instructblip: Towards general-purpose vision-language models with instruction tuning. Advances in Neural Information Processing Systems, 36.
  12. Qingxiu Dong, Lei Li, Damai Dai, Ce Zheng, Zhiyong Wu, Baobao Chang, Xu Sun, Jingjing Xu, and Zhifang Sui. 2022. A survey on in-context learning. arXiv preprint arXiv:2301.00234.
  13. Yushi Hu, Weijia Shi, Xingyu Fu, Dan Roth, Mari Ostendorf, Luke Zettlemoyer, Noah A Smith, and Ranjay Krishna. 2024. Visual sketchpad: Sketching as a visual chain of thought for multimodal language models. arXiv preprint arXiv:2406.09403. find
  14. Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexander C Berg, Wan-Yen Lo, and 1 others. 2023. Segment anything. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 4015–4026.
  15. Pan Lu, Hritik Bansal, Tony Xia, Jiacheng Liu, Chunyuan Li, Hannaneh Hajishirzi, Hao Cheng, Kai-Wei Chang, Michel Galley, and Jianfeng Gao. 2023. Mathvista: Evaluating mathematical reasoning of foundation models in visual contexts. arXiv preprint arXiv:2310.02255.
  16. Ning Miao, Yee Whye Teh, and Tom Rainforth. 2023. Selfcheck: Using llms to zero-shot check their own step-by-step reasoning. arXiv preprint arXiv:2308.00436. find
  17. Chancharik Mitra, Brandon Huang, Trevor Darrell, and Roei Herzig. 2023. Compositional chain-of-thought prompting for large multimodal models. arXiv preprint arXiv:2311.17076. find
  18. Chancharik Mitra, Brandon Huang, Trevor Darrell, and Roei Herzig. 2024. Compositional chain-of-thought prompting for large multimodal models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14420–14431. find
  19. Aleksandar Shtedritski, Christian Rupprecht, and Andrea Vedaldi. 2023. What does clip know about a red circle? visual prompt engineering for vlms. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 11987–11997. find
  20. Gemini Team, Rohan Anil, Sebastian Borgeaud, Yonghui Wu, Jean-Baptiste Alayrac, Jiahui Yu, Radu Soricut, Johan Schalkwyk, Andrew M Dai, Anja Hauth, and 1 others. 2023. Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805.
  21. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, and 1 others. 2023. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288.
  22. Xinlong Wang, Xiaosong Zhang, Yue Cao, Wen Wang, Chunhua Shen, and Tiejun Huang. 2023. Seggpt: Segmenting everything in context. arXiv preprint arXiv:2304.03284. — [arXiv:2304.03284](https://arxiv.org/abs/2304.03284) find
  23. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, and 1 others. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824– 24837.
  24. Chenfei Wu, Shengming Yin, Weizhen Qi, Xiaodong Wang, Zecheng Tang, and Nan Duan. 2023. Visual chatgpt: Talking, drawing and editing with visual foundation models. arXiv preprint arXiv:2303.04671.
  25. Deyao Zhu, Jun Chen, Xiaoqian Shen, Xiang Li, and Mohamed Elhoseiny. 2023. Minigpt-4: Enhancing vision-language understanding with advanced large language models. arXiv preprint arXiv:2304.10592.
  26. Xueyan Zou, Jianwei Yang, Hao Zhang, Feng Li, Linjie Li, Jianfeng Wang, Lijuan Wang, Jianfeng Gao, and Yong Jae Lee. 2024. Segment everything everywhere all at once. Advances in Neural Information Processing Systems, 36. find
  27. Jianwei Yang, Hao Zhang, Feng Li, Xueyan Zou, Chunyuan Li, and Jianfeng Gao. 2023a. Set-of-mark prompting unleashes extraordinary visual grounding in gpt-4v. arXiv preprint arXiv:2310.11441. find
  28. Zhengyuan Yang, Linjie Li, Kevin Lin, Jianfeng Wang, Chung-Ching Lin, Zicheng Liu, and Lijuan Wang. 2023b. The dawn of lmms: Preliminary explorations with gpt-4v (ision). arXiv preprint arXiv:2309.17421, 9(1):1.
  29. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. 2024. Tree of thoughts: Deliberate problem solving with large language models. Advances in Neural Information Processing Systems, 36. find
  30. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. 2022. React: Synergizing reasoning and acting in language models. arXiv preprint arXiv:2210.03629.
  31. Weihao Yu, Zhengyuan Yang, Linjie Li, Jianfeng Wang, Kevin Lin, Zicheng Liu, Xinchao Wang, and Lijuan Wang. 2023. Mm-vet: Evaluating large multimodal models for integrated capabilities. arXiv preprint arXiv:2308.02490.
  32. Xiang Yue, Yuansheng Ni, Kai Zhang, Tianyu Zheng, Ruoqi Liu, Ge Zhang, Samuel Stevens, Dongfu Jiang, Weiming Ren, Yuxuan Sun, and 1 others. 2023. Mmmu: A massive multi-discipline multimodal understanding and reasoning benchmark for expert agi. arXiv preprint arXiv:2311.16502.
  33. Renrui Zhang, Jiaming Han, Chris Liu, Peng Gao, Aojun Zhou, Xiangfei Hu, Shilin Yan, Pan Lu, Hongsheng Li, and Yu Qiao. 2023. Llama-adapter: Efficient fine-tuning of language models with zero-init attention. arXiv preprint arXiv:2303.16199.
  34. Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen, Christopher Dewan, Mona Diab, Xian Li, Xi Victoria Lin, and 1 others. 2022. Opt: Open pre-trained transformer language models. arXiv preprint arXiv:2205.01068.
  35. Huaixiu Steven Zheng, Swaroop Mishra, Xinyun Chen, Heng-Tze Cheng, Ed H Chi, Quoc V Le, and Denny Zhou. 2023. Take a step back: Evoking reasoning via abstraction in large language models. arXiv preprint arXiv:2310.06117. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-unac-adaptive-visual-prompting-abstraction/