---
title: "Visual Programming — paper by Gupta and Kembhavi (2022)"
description: "VisProg addresses complex, multi-step visual tasks described in natural language. 62.4% accuracy. 36 references, cited by 129 papers here."
authors: ["Tanmay Gupta", "Aniruddha Kembhavi"]
year: 2022
venue: "arXiv:2211.11559"
source: "https://arxiv.org/abs/2211.11559"
doi: "10.48550/arxiv.2211.11559"
arxiv: "2211.11559"
licence: "arXiv — non-exclusive license to distribute"
licence_url: "http://arxiv.org/licenses/nonexclusive-distrib/1.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/gupta-2022-visual-programming-compositional-visual-reasoning/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/gupta-2022-visual-programming-compositional-visual-reasoning/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2022](https://md.research.elamur.ai/free/en/papers/2022/page/2/) / Visual Programming: Compositional visual reasoning without training

# Visual Programming: Compositional visual reasoning without training

**Authors:** Tanmay Gupta and [Aniruddha Kembhavi](https://md.research.elamur.ai/free/en/authors/aniruddha-kembhavi/) · **Year:** 2022 · **Venue:** [arXiv:2211.11559](https://md.research.elamur.ai/free/en/venues/ieee-computer-society-conference-on-computer-vision-and-pattern-recognition/) ([IEEE Computer Society Conference on Computer Vision and Pattern Recognition 2022](https://md.research.elamur.ai/free/en/venues/ieee-computer-society-conference-on-computer-vision-and-pattern-recognition/2022/)) · **Publisher:** [Institute of Electrical and Electronics Engineers](https://md.research.elamur.ai/free/en/publishers/institute-of-electrical-and-electronics-engineers/) · **Licence:** [arXiv — non-exclusive license to distribute](http://arxiv.org/licenses/nonexclusive-distrib/1.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2211.11559) · [PDF](https://arxiv.org/pdf/2211.11559) · [DOI 10.48550/arxiv.2211.11559](https://doi.org/10.48550/arxiv.2211.11559) · [arXiv 2211.11559](https://arxiv.org/abs/2211.11559) · [OpenAlex](https://openalex.org/W4386065691)

**Citations:** 218 citations (OpenAlex, 2026-10-14) · [129](https://md.research.elamur.ai/free/en/papers/gupta-2022-visual-programming-compositional-visual-reasoning/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
VisProg addresses complex, multi-step visual tasks described in natural language. Collecting a dedicated training set for each such task does not scale. The system needs no gradient-based training. GPT-3 receives a few example pairs of instructions and programs, then writes a short Python-like program for a new instruction without seeing the image. Each program line calls one module from a library of about 20. These include pretrained neural models (OWL-ViT, DSFD, MaskFormer, CLIP, ViLT, Stable Diffusion), OpenCV-style image operations, and arithmetic or logical functions. Results from earlier lines feed later ones. Running the program produces both the answer and a visual rationale that exposes each intermediate output. The authors test it on compositional VQA, zero-shot reasoning over image pairs (NLVR), knowledge-based object tagging and language-guided image editing. It beats its base VQA model and performs strongly on tasks no single end-to-end model handles. This matters because new task types can be added through prompting alone, with transparent, debuggable reasoning.

## Key points
- On compositional VQA, VisProg improves on the underlying ViLT VQA model by 2.7 points. On NLVRv2 it reaches 62.4% accuracy zero-shot, using only a single-image VQA model and never training on image pairs. A ViLT model fine-tuned on NLVRv2 serves as the upper bound.
- On knowledge tagging, it scores 63.7% F1 for tagging (correct box plus correct name) and 80.6% F1 for localization alone. Here GPT-3 also acts as the knowledge source.
- The authors inspected rationales for roughly 100 samples per task. Wrong programs were the main error source on GQA, affecting 16% of samples. A stronger VQA module could improve NLVR results by up to 24%.
- Users can read the rationales and rephrase instructions, for example by adding a category or a list size. This "instruction tuning" clearly improved the tagging and editing results.

*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

- [Allen Institute for Artificial Intelligence](https://md.research.elamur.ai/free/en/labs/allen-institute-for-artificial-intelligence/) — Research institute, Seattle, United States · 304 works · 106,137 citations

📚 Shelves (3): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 3,486 works

🧭 Topics (1): [🧮 Reasoning](https://md.research.elamur.ai/free/en/topics/reasoning/) · 1,685 works

## Cite

Cite as: Tanmay Gupta and Aniruddha Kembhavi. 2022. Visual Programming: Compositional visual reasoning without training. arXiv:2211.11559.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{gupta2022visual,
  title = {{Visual Programming: Compositional visual reasoning without training}},
  author = {Tanmay Gupta and Aniruddha Kembhavi},
  year = {2022},
  doi = {10.48550/arxiv.2211.11559},
  eprint = {2211.11559},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2211.11559}
}
```

- [← previous UL2: Unifying Language Learning Paradigms](https://md.research.elamur.ai/free/en/papers/tay-2022-ul2-unifying-language-learning-paradigms/)
- [next → Transformer Feed-Forward Layers Build Predictions by Promoting Concepts in the Vocabulary Space](https://md.research.elamur.ai/free/en/papers/geva-2022-transformer-feed-forward-layers-build/)
- [↑ back to papers of 2022](https://md.research.elamur.ai/free/en/papers/2022/page/2/)

## References

36 references, 18 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Yana Hasson, Karel Lenc, Arthur Mensch, Katie Millican, Malcolm Reynolds, Roman Ring, Eliza Rutherford, Serkan Cabi, Tengda Han, Zhitao Gong, Sina Samangooei, Marianne Monteiro, Jacob Menick, Sebastian Borgeaud, Andy Brock,… — [Flamingo: a Visual Language Model for Few-Shot Learning](https://md.research.elamur.ai/free/en/papers/alayrac-2022-flamingo-visual-language-model-few/)
2. Jacob Andreas, Marcus Rohrbach, Trevor Darrell, and Dan Klein. [Neural module networks](https://md.research.elamur.ai/free/en/papers/andreas-2015-neural-module-networks/). 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 39–48, 2016. 2, 3
3. Ankur Bapna, Colin Cherry, Yu Zhang, Ye Jia, Melvin Johnson, Yong Cheng, Simran Khanuja, Jason Riesa, and Alexis Conneau. mslam: Massively multilingual joint pre-training for speech and text. ArXiv, abs/2202.01374, 2022. 3 [find](https://scholar.google.com/scholar?q=Ankur+Bapna%2C+Colin+Cherry%2C+Yu+Zhang%2C+Ye+Jia%2C+Melvin+Johnson%2C+Yong+Cheng%2C+Simran+Khanuja%2C+Jason+Riesa%2C+and+Alexis+Conneau%2E+mslam%3A+Massively+multilingual+joint+pre%2Dtraining+for+speech+and+text%2E+ArXiv%2C+a)
4. Gary Bradski. The opencv library. Dr. Dobb’s Journal: Software Tools for the Professional Programmer, 25(11):120– 123, 2000. 2 [find](https://scholar.google.com/scholar?q=Gary+Bradski%2E+The+opencv+library%2E+Dr%2E+Dobb%E2%80%99s+Journal%3A+Software+Tools+for+the+Professional+Programmer%2C+25%2811%29%3A120%E2%80%93+123%2C+2000%2E+2)
5. Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, T. J. Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeff Wu, Clemens Winter, C… — [Language Models are Few-Shot Learners](https://md.research.elamur.ai/free/en/papers/brown-2020-language-models-few-shot-learners/)
6. Bowen Cheng, Alexander G. Schwing, and Alexander Kirillov. Per-pixel classification is not all you need for semantic segmentation. In NeurIPS, 2021. 2 [find](https://scholar.google.com/scholar?q=Bowen+Cheng%2C+Alexander+G%2E+Schwing%2C+and+Alexander+Kirillov%2E+Per%2Dpixel+classification+is+not+all+you+need+for+semantic+segmentation%2E+In+NeurIPS%2C+2021%2E+2)
7. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. [Bert: Pre-training of deep bidirectional transformers for language understanding](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/). ArXiv, abs/1810.04805, 2019. 3
8. Tanmay Gupta, Amita Kamath, Aniruddha Kembhavi, and Derek Hoiem. Towards general purpose vision systems. ArXiv, abs/2104.00743, 2021. 2 — [arXiv:2104.00743](https://arxiv.org/abs/2104.00743) [find](https://scholar.google.com/scholar?q=Tanmay+Gupta%2C+Amita+Kamath%2C+Aniruddha+Kembhavi%2C+and+Derek+Hoiem%2E+Towards+general+purpose+vision+systems%2E+ArXiv%2C+abs%2F2104%2E00743%2C+2021%2E+2+%E2%80%94+%5BarXiv%3A2104%2E00743%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2104%2E00743%29)
9. Ronghang Hu, Jacob Andreas, Trevor Darrell, and Kate Saenko. Explainable neural computation via stack neural module networks. In ECCV, 2018. 2, 3 [find](https://scholar.google.com/scholar?q=Ronghang+Hu%2C+Jacob+Andreas%2C+Trevor+Darrell%2C+and+Kate+Saenko%2E+Explainable+neural+computation+via+stack+neural+module+networks%2E+In+ECCV%2C+2018%2E+2%2C+3)
10. Ronghang Hu, Jacob Andreas, Marcus Rohrbach, Trevor Darrell, and Kate Saenko. Learning to reason: End-to-end module networks for visual question answering. 2017 IEEE International Conference on Computer Vision (ICCV), pages 804–813, 2017. 2, 3 [find](https://scholar.google.com/scholar?q=Ronghang+Hu%2C+Jacob+Andreas%2C+Marcus+Rohrbach%2C+Trevor+Darrell%2C+and+Kate+Saenko%2E+Learning+to+reason%3A+End%2Dto%2Dend+module+networks+for+visual+question+answering%2E+2017+IEEE+International+Conference+on+Comput)
11. Drew A. Hudson and Christopher D. Manning. [Gqa: A new dataset for real-world visual reasoning and compositional question answering](https://md.research.elamur.ai/free/en/papers/hudson-2019-gqa-new-dataset-real-world/). 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 6693– 6702, 2019. 5
12. Justin Johnson, Bharath Hariharan, Laurens van der Maaten, Judy Hoffman, Li Fei-Fei, C. Lawrence Zitnick, and Ross B. Girshick. [Inferring and executing programs for visual reasoning](https://md.research.elamur.ai/free/en/papers/johnson-2017-inferring-executing-programs-visual-reasoning/). 2017 IEEE International Conference on Computer Vision (ICCV), pages 3008–3017, 2017. 2, 3
13. Amita Kamath, Christopher Clark, Tanmay Gupta, Eric Kolve, Derek Hoiem, and Aniruddha Kembhavi. Webly supervised concept expansion for general purpose vision models. In ECCV, 2022. 2 [find](https://scholar.google.com/scholar?q=Amita+Kamath%2C+Christopher+Clark%2C+Tanmay+Gupta%2C+Eric+Kolve%2C+Derek+Hoiem%2C+and+Aniruddha+Kembhavi%2E+Webly+supervised+concept+expansion+for+general+purpose+vision+models%2E+In+ECCV%2C+2022%2E+2)
14. Tushar Khot, Kyle Richardson, Daniel Khashabi, and Ashish Sabharwal. Learning to solve complex tasks by talking to agents. ArXiv, abs/2110.08542, 2021. 3 — [arXiv:2110.08542](https://arxiv.org/abs/2110.08542) [find](https://scholar.google.com/scholar?q=Tushar+Khot%2C+Kyle+Richardson%2C+Daniel+Khashabi%2C+and+Ashish+Sabharwal%2E+Learning+to+solve+complex+tasks+by+talking+to+agents%2E+ArXiv%2C+abs%2F2110%2E08542%2C+2021%2E+3+%E2%80%94+%5BarXiv%3A2110%2E08542%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F211)
15. Tushar Khot, H. Trivedi, Matthew Finlayson, Yao Fu, Kyle Richardson, Peter Clark, and Ashish Sabharwal. [Decomposed prompting: A modular approach for solving complex tasks](https://md.research.elamur.ai/free/en/papers/khot-2022-decomposed-prompting-modular-approach/). ArXiv, abs/2210.02406, 2022. 3
16. Wonjae Kim, Bokyung Son, and Ildoo Kim. [Vilt: Vision-and-language transformer without convolution or region supervision](https://md.research.elamur.ai/free/en/papers/kim-2021-vilt-vision-language-transformer-convolution/). In ICML, 2021. 2, 5
17. Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. [Large language models are zero-shot reasoners](https://md.research.elamur.ai/free/en/papers/kojima-2022-large-language-models-zero-shot/). ArXiv, abs/2205.11916, 2022. 3
18. Jian Li, Yabiao Wang, Changan Wang, Ying Tai, Jianjun Qian, Jian Yang, Chengjie Wang, Jilin Li, and Feiyue Huang. Dsfd: Dual shot face detector. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 5055–5064, 2019. 2 [find](https://scholar.google.com/scholar?q=Jian+Li%2C+Yabiao+Wang%2C+Changan+Wang%2C+Ying+Tai%2C+Jianjun+Qian%2C+Jian+Yang%2C+Chengjie+Wang%2C+Jilin+Li%2C+and+Feiyue+Huang%2E+Dsfd%3A+Dual+shot+face+detector%2E+2019+IEEE%2FCVF+Conference+on+Computer+Vision+and+Pattern)
19. Jiasen Lu, Christopher Clark, Rowan Zellers, Roozbeh Mottaghi, and Aniruddha Kembhavi. Unified-io: A unified model for vision, language, and multi-modal tasks. ArXiv, abs/2206.08916, 2022. 2 — [arXiv:2206.08916](https://arxiv.org/abs/2206.08916) [find](https://scholar.google.com/scholar?q=Jiasen+Lu%2C+Christopher+Clark%2C+Rowan+Zellers%2C+Roozbeh+Mottaghi%2C+and+Aniruddha+Kembhavi%2E+Unified%2Dio%3A+A+unified+model+for+vision%2C+language%2C+and+multi%2Dmodal+tasks%2E+ArXiv%2C+abs%2F2206%2E08916%2C+2022%2E+2+%E2%80%94+%5BarXiv%3A)
20. Kenneth Marino, Mohammad Rastegari, Ali Farhadi, and Roozbeh Mottaghi. Ok-vqa: A visual question answering benchmark requiring external knowledge. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 3190–3199, 2019. 3 [find](https://scholar.google.com/scholar?q=Kenneth+Marino%2C+Mohammad+Rastegari%2C+Ali+Farhadi%2C+and+Roozbeh+Mottaghi%2E+Ok%2Dvqa%3A+A+visual+question+answering+benchmark+requiring+external+knowledge%2E+2019+IEEE%2FCVF+Conference+on+Computer+Vision+and+Patte)
21. Matthias Minderer, Alexey A. Gritsenko, Austin Stone, Maxim Neumann, Dirk Weissenborn, Alexey Dosovitskiy, Aravindh Mahendran, Anurag Arnab, Mostafa Dehghani, Zhuoran Shen, Xiao Wang, Xiaohua Zhai, Thomas Kipf, and Neil Houlsby. Simple open-vocabulary object detection with vision transformers. ArXiv… — [arXiv:2205.06230](https://arxiv.org/abs/2205.06230) [find](https://scholar.google.com/scholar?q=Matthias+Minderer%2C+Alexey+A%2E+Gritsenko%2C+Austin+Stone%2C+Maxim+Neumann%2C+Dirk+Weissenborn%2C+Alexey+Dosovitskiy%2C+Aravindh+Mahendran%2C+Anurag+Arnab%2C+Mostafa+Dehghani%2C+Zhuoran+Shen%2C+Xiao+Wang%2C+Xiaohua+Zhai%2C+Th)
22. Zoe Papakipos and Joanna Bitton. Augly: Data augmentations for robustness. ArXiv, abs/2201.06494, 2022. 12 — [arXiv:2201.06494](https://arxiv.org/abs/2201.06494) [find](https://scholar.google.com/scholar?q=Zoe+Papakipos+and+Joanna+Bitton%2E+Augly%3A+Data+augmentations+for+robustness%2E+ArXiv%2C+abs%2F2201%2E06494%2C+2022%2E+12+%E2%80%94+%5BarXiv%3A2201%2E06494%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2201%2E06494%29)
23. Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. [Learning transferable visual models from natural language supervision](https://md.research.elamur.ai/free/en/papers/radford-2021-learning-transferable-visual-models-natural/). In International Conference on Machine Learning, pages 8748–8763. PMLR, 20…
24. Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. Language models are unsupervised multitask learners. 2019. 3 [find](https://scholar.google.com/scholar?q=Alec+Radford%2C+Jeff+Wu%2C+Rewon+Child%2C+David+Luan%2C+Dario+Amodei%2C+and+Ilya+Sutskever%2E+Language+models+are+unsupervised+multitask+learners%2E+2019%2E+3)
25. Colin Raffel, Noam M. Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. [Exploring the limits of transfer learning with a unified text-to-text transformer](https://md.research.elamur.ai/free/en/papers/raffel-2019-exploring-limits-transfer-learning-unified/). ArXiv, abs/1910.10683, 2020. 2
26. Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea Voss, Alec Radford, Mark Chen, and Ilya Sutskever. [Zero-shot text-to-image generation](https://md.research.elamur.ai/free/en/papers/ramesh-2021-zero-shot-text-image-generation/). ArXiv, abs/2102.12092, 2021. 7
27. Scott Reed, Konrad Zolna, Emilio Parisotto, Sergio Gomez Colmenarejo, Alexander Novikov, Gabriel Barth-Maron, Mai Gimenez, Yury Sulsky, Jackie Kay, Jost Tobias Springenberg, Tom Eccles, Jake Bruce, Ali Razavi, Ashley D. Edwards, Nicolas Manfred Otto Heess, Yutian Chen, Raia Hadsell, Oriol Vinyals, M… [find](https://scholar.google.com/scholar?q=Scott+Reed%2C+Konrad+Zolna%2C+Emilio+Parisotto%2C+Sergio+Gomez+Colmenarejo%2C+Alexander+Novikov%2C+Gabriel+Barth%2DMaron%2C+Mai+Gimenez%2C+Yury+Sulsky%2C+Jackie+Kay%2C+Jost+Tobias+Springenberg%2C+Tom+Eccles%2C+Jake+Bruce%2C+Al)
28. Robin Rombach, A. Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. [High-resolution image synthesis with latent diffusion models](https://md.research.elamur.ai/free/en/papers/rombach-2021-high-resolution-image-synthesis-latent/). 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 10674–10685, 2022. 2, 7
29. Ishika Singh, Valts Blukis, Arsalan Mousavian, Ankit Goyal, Danfei Xu, Jonathan Tremblay, Dieter Fox, Jesse Thomason, and Animesh Garg. [Progprompt: Generating situated robot task plans using large language models](https://md.research.elamur.ai/free/en/papers/singh-2022-progprompt-generating-situated-robot-task/). ArXiv, abs/2209.11302, 2022. 3
30. Alane Suhr, Stephanie Zhou, Iris Zhang, Huajun Bai, and Yoav Artzi. [A corpus for reasoning about natural language grounded in photographs](https://md.research.elamur.ai/free/en/papers/suhr-2019-corpus-reasoning-natural-language-grounded/). ArXiv, abs/1811.00491, 2019. 6
31. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, and Denny Zhou. [Self-consistency improves chain of thought reasoning in language models](https://md.research.elamur.ai/free/en/papers/wang-2022-self-consistency-improves-chain-thought/). ArXiv, abs/2203.11171, 2022. 7
32. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed Chi, Quoc Le, and Denny Zhou. Chain of thought prompting elicits reasoning in large language models. ArXiv, abs/2201.[11903, 2022](https://md.research.elamur.ai/free/en/papers/wei-2022-chain-thought-prompting-elicits-reasoning/). 3, 7
33. Ronald J Williams. Simple statistical gradient-following algorithms for connectionist reinforcement learning. Machine learning, 8(3):229–256, 1992. 2, 3 [find](https://scholar.google.com/scholar?q=Ronald+J+Williams%2E+Simple+statistical+gradient%2Dfollowing+algorithms+for+connectionist+reinforcement+learning%2E+Machine+learning%2C+8%283%29%3A229%E2%80%93256%2C+1992%2E+2%2C+3)
34. Zhengyuan Yang, Zhe Gan, Jianfeng Wang, Xiaowei Hu, Yumao Lu, Zicheng Liu, and Lijuan Wang. An empirical study of gpt-3 for few-shot knowledge-based vqa. In AAAI, 2022. 3 [find](https://scholar.google.com/scholar?q=Zhengyuan+Yang%2C+Zhe+Gan%2C+Jianfeng+Wang%2C+Xiaowei+Hu%2C+Yumao+Lu%2C+Zicheng+Liu%2C+and+Lijuan+Wang%2E+An+empirical+study+of+gpt%2D3+for+few%2Dshot+knowledge%2Dbased+vqa%2E+In+AAAI%2C+2022%2E+3)
35. Jiahui Yu, Yuanzhong Xu, Jing Yu Koh, Thang Luong, Gunjan Baid, Zirui Wang, Vijay Vasudevan, Alexander Ku, Yinfei Yang, Burcu Karagol Ayan, Benton C. Hutchinson, Wei Han, Zarana Parekh, Xin Li, Han Zhang, Jason Baldridge, and Yonghui Wu. Scaling autoregressive models for content-rich text-to-image g… — [arXiv:2206.10789](https://arxiv.org/abs/2206.10789) [find](https://scholar.google.com/scholar?q=Jiahui+Yu%2C+Yuanzhong+Xu%2C+Jing+Yu+Koh%2C+Thang+Luong%2C+Gunjan+Baid%2C+Zirui+Wang%2C+Vijay+Vasudevan%2C+Alexander+Ku%2C+Yinfei+Yang%2C+Burcu+Karagol+Ayan%2C+Benton+C%2E+Hutchinson%2C+Wei+Han%2C+Zarana+Parekh%2C+Xin+Li%2C+Han+Zh)
36. Andy Zeng, Maria Attarian, Brian Ichter, Krzysztof Choromanski, Adrian Wong, Stefan Welker, Federico Tombari, Aveek Purohit, Michael Ryoo, Vikas Sindhwani, Johnny Lee, Vincent Vanhoucke, and Pete Florence. [Socratic models: Composing zero-shot multimodal reasoning with language](https://md.research.elamur.ai/free/en/papers/zeng-2022-socratic-models-composing-zero-shot/). arXiv, 2022. 3

*Updated 2 November 2026*