Elamur Research⌕
☰
home / papers / 2026 / DeepPresenter: Environment-Grounded Reflection for Agentic Presentation Generation

DeepPresenter: Environment-Grounded Reflection for Agentic Presentation Generation

Paper · 2026 · Hao Zheng, Guozhao Mo, Xinru Yan et al.

Summary

Building slide decks automatically requires solid research, a visual style that fits the topic, and repeated correction. Existing agents often fall short because they follow fixed pipelines and generic templates. They also check their own code or reasoning, which cannot reveal problems that only appear once a slide is rendered.

DeepPresenter addresses this with two agents that share one backbone model and communicate through a common file system:

  • Researcher explores sources and writes a structured markdown manuscript.
  • Presenter creates a design plan and builds each slide as HTML from scratch.

Reflection is tied to what the environment shows. An inspect tool renders slides into images or reports the manuscript's status, and the agent then plans targeted fixes.

The authors also distilled a compact model, DeepPresenter-9B, from synthesized trajectories. During synthesis, an independent critic judged the outputs to reduce self-verification bias.

On 128 held-out tasks, DeepPresenter outperformed open-source systems and the commercial tool Gamma. The 9B model came close to GPT-5 at much lower cost, which makes good automatic presentation generation more affordable.

Key points

  • Task collection: 1,152 tasks, of which 1,024 were used for trajectory synthesis and 128 held out for evaluation.
    • Sources: PersonaHub, arXiv and FinePDFs-Edu.
    • Languages: 603 English and 549 Chinese.
    • Each task has checkable constraints (slide count, language, aspect ratio). GLM-4.6 generated the persona-based tasks.
  • Evaluation: systems were scored on constraint satisfaction, content quality and visual style.
    • DeepPresenter with proprietary backbones averaged 4.44, ahead of Gamma at 4.36.
    • Its scores ranged from 3.82 to 4.44 depending on the backbone, while baseline frameworks stayed roughly flat across backbones.
  • Cost-efficiency: DeepPresenter-9B scored 4.19, versus GPT-5 at 4.22. At similar cost it clearly beat KCTV + Gemini-3-Pro, which scored 3.83.
  • Failure analysis:
    • Environment failures (context overflow, infrastructure issues): 32.3%.
    • Constraint violations: 13.5%.
    • Consistency errors: 11.2%.
  • Limitations: the external critic is used only during training-data synthesis, not at inference time. Multi-step tool rollouts are also costly and fragile. Code is available at github.com/icip-cas/PPTAgent.

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

🏛️Chinese Academy of SciencesBeijing · China246 works · 24,952 citations🎓University of Chinese Academy of SciencesBeijing · China188 works · 8,744 citations🔬Institute of SoftwareBeijing · China27 works · 4,980 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

Cite

Cite as Hao Zheng, Guozhao Mo, Xinru Yan, Qianhao Yuan, Wenkai Zhang, Xuanang Chen, Yaojie Lu, Hongyu Lin, Xianpei Han and Le Sun. 2026. DeepPresenter: Environment-Grounded Reflection for Agentic Presentation Generation. 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.1578

@inproceedings{zheng2026deeppresenter,
  title = {{DeepPresenter: Environment-Grounded Reflection for Agentic Presentation Generation}},
  author = {Hao Zheng, Guozhao Mo, Xinru Yan, Qianhao Yuan, Wenkai Zhang, Xuanang Chen, Yaojie Lu, Hongyu Lin, Xianpei Han and Le Sun},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1578},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1578}
}

References

32 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Anthropic. 2025. Introducing claude sonnet 4.5. https://www.anthropic.com/news/ claude-sonnet-4-5. [Accessed 18-11-2025]. ↗
  2. Isabel Alyssa Cachola, Silviu Cucerzan, Allen Herring, Vuksan Mijovic, Erik Oveson, and Sujay Kumar Jauhar. 2024. Knowledge-centric templatic views of documents. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 15460–15476, Miami, Florida, USA. Association for Computat… find
  3. Gheorghe Comanici, Eric Bieber, Mike Schaekermann, Ice Pasupat, Noveen Sachdeva, Inderjit Dhillon, Marcel Blistein, Ori Ram, Dan Zhang, Evan Rosen, Luke Marris, Sam Petulla, Colin Gaffney, Asaf Aharoni, Nathan Lintz, Tiago Cardal Pais, Henrik Jacobsson, Idan Szpektor, Nan-Jiang Jiang, and others. 20… — Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities
  4. Jiaxin Ge, Zora Zhiruo Wang, Xuhui Zhou, Yi-Hao Peng, Sanjay Subramanian, Qinyue Tan, Maarten Sap, Alane Suhr, Daniel Fried, Graham Neubig, and others. 2025. Autopresent: Designing structured visuals from scratch. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 2902–2… find
  5. Tao Ge, Xin Chan, Xiaoyang Wang, Dian Yu, Haitao Mi, and Dong Yu. 2024. Scaling synthetic data creation with 1,000,000,000 personas. arXiv preprint arXiv:2406.20094.
  6. Yancheng He, Shilong Li, Jiaheng Liu, Weixun Wang, Xingyuan Bu, Ge Zhang, Z.y. Peng, Zhaoxiang Zhang, Zhicheng Zheng, Wenbo Su, and Bo Zheng. 2025. Can large language models detect errors in long chain-of-thought reasoning? In Proceedings of the 63rd Annual Meeting of the Association for Computation…
  7. Yue Hu and Xiaojun Wan. 2014. Ppsgen: Learning-based presentation slides generation for academic papers. IEEE transactions on knowledge and data engineering, 27(4):1085–1097. find
  8. Zheng Huang, Xukai Liu, Tianyu Hu, Kai Zhang, and Ye Liu. 2025. Pptbench: Towards holistic evaluation of large language models for powerpoint layout and design understanding. arXiv preprint arXiv:2512.02624. — [arXiv:2512.02624](https://arxiv.org/abs/2512.02624) find
  9. Dongwei Jiang, Jingyu Zhang, Orion Weller, Nathaniel Weir, Benjamin Van Durme, and Daniel Khashabi. 2025. Self-[in] correct: Llms struggle with discriminating self-generated responses. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 24266–24275. find
  10. Kyudan Jung, Hojun Cho, Jooyeol Yun, Soyoung Yang, Jaehyeok Jang, and Jaegul Choo. 2025. Talk to your slides: Language-driven agents for efficient slide editing. arXiv preprint arXiv:2505.11604. — [arXiv:2505.11604](https://arxiv.org/abs/2505.11604) find
  11. Jeonghye Kim, Sojeong Rhee, Minbeom Kim, Dohyung Kim, Sangmook Lee, Youngchul Sung, and Kyomin Jung. 2025. Reflact: World-grounded decision making in llm agents via goal-state reflection. arXiv preprint arXiv:2505.15182. find
  12. Hynek Kydlı́ček, Guilherme Penedo, and Leandro von Werra. 2025. Finepdfs. https://huggingface.co/ datasets/HuggingFaceFW/finepdfs edu. ↗
  13. Dan Friedman and Adji Bousso Dieng. 2022. The vendi score: A diversity evaluation metric for machine learning. arXiv preprint arXiv:2210.02410. — [arXiv:2210.02410](https://arxiv.org/abs/2210.02410) find
  14. Xin Liang, Xiang Zhang, Yiwei Xu, Siqi Sun, and Chenyu You. 2025. Slidegen: Collaborative multimodal agents for scientific slide generation. arXiv preprint arXiv:2512.04529. — [arXiv:2512.04529](https://arxiv.org/abs/2512.04529) find
  15. Tsu-Jui Fu, William Yang Wang, Daniel McDuff, and Yale Song. 2022. Doc2ppt: Automatic presentation slides generation from scientific documents. Proceedings of the AAAI Conference on Artificial Intelligence, 36(1):634–642. find
  16. Chengzhi Liu, Yuzhe Yang, Kaiwen Zhou, Zhen Zhang, Yue Fan, Yanan Xie, Peng Qi, and Xin Eric Wang. 2025. Presenting a paper is an art: Self-improvement aesthetic agents for academic presentations. arXiv preprint arXiv:2510.05571. — [arXiv:2510.05571](https://arxiv.org/abs/2510.05571) find
  17. OpenAI. 2025. Introducing gpt-5. https://openai.com/index/introducing-gpt-5/. [Accessed 18-11-2025]. ↗
  18. Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy Vo, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, and others. 2023. Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193.
  19. Qianhao Yuan, Jie Lou, Zichao Li, Jiawei Chen, Yaojie Lu, Hongyu Lin, Le Sun, Debing Zhang, and Xianpei Han. 2025. Memsearcher: Training llms to reason, search and manage memory via end-to-end reinforcement learning. arXiv preprint arXiv:2511.02805. find
  20. Aohan Zeng, Xin Lv, Qinkai Zheng, Zhenyu Hou, Bin Chen, Chengxing Xie, Cunxiang Wang, Da Yin, Hao Zeng, Jiajie Zhang, and others. 2025a. Glm-4.5: Agentic, reasoning, and coding (arc) foundation models. arXiv preprint arXiv:2508.06471.
  21. Athar Sefid, Prasenjit Mitra, and Lee Giles. 2021. Slidegen: an abstractive section-based slide generator for scholarly documents. In Proceedings of the 21st ACM Symposium on Document Engineering, pages 1–4. find
  22. Wenzheng Zeng, Mingyu Ouyang, Langyuan Cui, and Hwee Tou Ng. 2025b. Slidetailor: Personalized presentation slide generation for scientific papers. arXiv preprint arXiv:2512.20292. — [arXiv:2512.20292](https://arxiv.org/abs/2512.20292) find
  23. Kaya Stechly, Karthik Valmeekam, and Subbarao Kambhampati. 2024. On the self-verification limitations of large language models on reasoning and planning tasks. arXiv preprint arXiv:2402.08115. find
  24. Yuze Zhao, Jintao Huang, Jinghan Hu, Xingjun Wang, Yunlin Mao, Daoze Zhang, Zeyinzi Jiang, Zhikai Wu, Baole Ai, Ang Wang, Wenmeng Zhou, and Yingda Chen. 2024. Swift:a scalable lightweight infrastructure for fine-tuning. Preprint, arXiv:2408.05517. — [arXiv:2408.05517](https://arxiv.org/abs/2408.05517) find
  25. Edward Sun, Yufang Hou, Dakuo Wang, Yunfeng Zhang, and Nancy XR Wang. 2021. D2s: Document-to-slide generation via query-based text summarization. arXiv preprint arXiv:2105.03664. find
  26. Zhengyang Tang, Ziniu Li, Zhenyang Xiao, Tian Ding, Ruoyu Sun, Benyou Wang, Dayiheng Liu, Fei Huang, Tianyu Liu, Bowen Yu, and others. 2025. Realcritic: Towards effectiveness-driven evaluation of language model critiques. arXiv preprint arXiv:2501.14492. — [arXiv:2501.14492](https://arxiv.org/abs/2501.14492) find
  27. Hao Zheng, Xinyan Guan, Hao Kong, Wenkai Zhang, Jia Zheng, Weixiang Zhou, Hongyu Lin, Yaojie Lu, Xianpei Han, and Le Sun. 2025. Pptagent: Generating and evaluating presentations beyond text-to-slides. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 14…
  28. V Team, Wenyi Hong, Wenmeng Yu, Xiaotao Gu, Guo Wang, Guobing Gan, Haomiao Tang, Jiale Cheng, Ji Qi, Junhui Ji, Lihang Pan, Shuaiqi Duan, Weihan Wang, Yan Wang, Yean Cheng, Zehai He, Zhe Su, Zhen Yang, Ziyang Pan, and others. 2025. Glm-4.5v and glm-4.1v-thinking: Towards versatile multimodal reasoni…
  29. Sida Wang, Xiaojun Wan, and Shikang Du. 2017. Phrase-based presentation slides generation for academic papers. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 31. find
  30. Xiaojie Xu, Xinli Xu, Sirui Chen, Haoyu Chen, Fan Zhang, and Ying-Cong Chen. 2025. Pregenie: An agentic framework for high-quality visual presentation generation. arXiv preprint arXiv:2505.21660. find
  31. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, and others. 2025a. Qwen3 technical report. arXiv preprint arXiv:2505.09388.
  32. Yuheng Yang, Wenjia Jiang, Yang Wang, Yiwei Wang, and Chi Zhang. 2025b. Auto-slides: An interactive multi-agent system for creating and customizing research presentations. arXiv preprint arXiv:2509.11062. — [arXiv:2509.11062](https://arxiv.org/abs/2509.11062) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zheng-2026-deeppresenter-environment-grounded-reflection-agentic/