Elamur Research⌕
☰
home / papers / 2026 / CogGen: A Cognitively Inspired Recursive Framework for Deep Research Report Generation

CogGen: A Cognitively Inspired Recursive Framework for Deep Research Report Generation

Paper · 2026 · Kuo Tian, Pengfei Sun, Zhen Wu et al.

Summary

Agent pipelines that write long research reports with LLMs usually run in one direction: they draft an outline, then fill it in order. Early mistakes therefore carry forward, the structure cannot be revised when later findings call for it, and charts tend to be produced apart from the prose. CogGen borrows from cognitive theories of writing, where planning, drafting and reviewing alternate. Its two-level recursive design pairs a Planner, a Writer and a Reviewer agent. They repeat a global cycle that can reorganise the whole outline and earlier text, and each section has its own search–replan–write cycle. Charts are planned through Abstract Visual Representation (AVR), a compact way to state what a chart should show; a separate Render Agent draws it later. The authors also introduce CLEF, an evaluation scheme based on cognitive load theory, and a new benchmark built from Our World in Data (OWID). CogGen leads the open-source systems tested and comes close to professional analysts. This suggests revisable, non-linear agent designs can improve automated analytical writing.

Key points

  • Architecture: a Macro-Cognitive Loop plans, writes and reviews the full draft, then updates the plan using the Reviewer's feedback. A Micro-Cognitive Cycle runs section drafts in parallel threads, and these are merged afterwards.
  • AVR vs. FDV: in MMDR, the Writer must produce the full Formal Description of Visualization. AVR captures only what the chart is meant to convey, which reduces dual-task load on the Writer.
  • Chart hallucination: chart-data hallucination rates were 60% for FDV (MMDR), 67% for AVR without checks, and 28% for AVR with a Post-Rendering Audit. The audit checks the rendered data against the knowledge base.
  • CLEF dimensions: Organization, Depth, Relevance, Alignment and Synergy. Evaluation used the OWID benchmark and WildSeek; on WildSeek, CogGen beat reference reports from Gemini Deep Research.
  • Limitations: the recursive loops add compute cost, the automatic charts still fall short of expert-made ones, and rendering is restricted to ECharts and Mermaid. Code and data are released at github.com/NJUNLP/CogGen.

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

🎓Nanjing UniversityNanjing · China74 works · 7,814 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

Cite

Cite as Kuo Tian, Pengfei Sun, Zhen Wu, Junran Ding and Xinyu Dai. 2026. CogGen: A Cognitively Inspired Recursive Framework for Deep Research Report 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.296

@inproceedings{tian2026coggen,
  title = {{CogGen: A Cognitively Inspired Recursive Framework for Deep Research Report Generation}},
  author = {Kuo Tian, Pengfei Sun, Zhen Wu, Junran Ding and Xinyu Dai},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.296},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.296}
}

References

35 references, 10 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Mingxuan Du, Benfeng Xu, Chiwei Zhu, Xiaorui Wang, and Zhendong Mao. 2025. DeepResearch Bench: A Comprehensive Benchmark for Deep Research Agents. Preprint, arXiv:2506.11763.
  2. Linda Flower and John R. Hayes. 1981. A Cognitive Process Theory of Writing. College Composition and Communication, 32(4):365–387. find
  3. Alireza Ghafarollahi and Markus J. Buehler. 2025. Sci-Agents: Automating Scientific Discovery Through Bioinspired Multi-Agent Intelligent Graph Reasoning. Advanced Materials, 37(22):2413523. find
  4. Google. 2025. Gemini deep research — your personal research assistant. Technical report, Google. https://gemini.google/overview/deep-research/. ↗
  5. Baixuan Li, Bo Zhang, Dingchu Zhang, Fei Huang, Guangyu Li, Guoxin Chen, Huifeng Yin, Jialong Wu, Jingren Zhou, Kuan Li, Liangcai Su, Litu Ou, Liwen Zhang, Pengjun Xie, Rui Ye, Wenbiao Yin, Xinmiao Yu, Xinyu Wang, Xixi Wu, and 37 others. 2025a. Tongyi deepresearch technical report. Preprint, arXiv:2… — [arXiv:2510.24701](https://arxiv.org/abs/2510.24701) find
  6. Deqing Li, Honghui Mei, Yi Shen, Shuang Su, Wenli Zhang, Junting Wang, Ming Zu, and Wei Chen. 2018. ECharts: A declarative framework for rapid construction of web-based visualization. Visual Informatics, 2(2):136–146. find
  7. Xiaoxi Li, Guanting Dong, Jiajie Jin, Yuyao Zhang, Yujia Zhou, Yutao Zhu, Peitian Zhang, and Zhicheng Dou. 2025b. Search-o1: Agentic Search-Enhanced Large Reasoning Models. Preprint, arXiv:2501.05366.
  8. Xinze Li, Zhenghao Liu, Haidong Xin, Yukun Yan, Shuo Wang, Zheni Zeng, Sen Mei, Ge Yu, and Maosong Sun. 2026. Structured Knowledge Representation through Contextual Pages for Retrieval-Augmented Generation. Preprint, arXiv:2601.09402. find
  9. Chin-Yew Lin. 2004. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out, pages 74–81.
  10. Nelson F Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. 2024. Lost in the middle: How language models use long contexts. Transactions of the Association for Computational Linguistics, 12:157–173.
  11. Richard E Mayer. 2005. Cognitive theory of multimedia learning. The Cambridge handbook of multimedia learning, 41(1):31–48. find
  12. OpenAI. 2025a. GPT-4.1. Technical report, OpenAI. https://openai.com/index/gpt-4-1/. ↗
  13. Knut Sveidqvist and Mermaid Development Team. 2014. Mermaid: Generation of diagrams and flowcharts from text. Software, MIT License. find
  14. John Sweller. 1994. Cognitive load theory, learning difficulty, and instructional design. Learning and instruction, 4(4):295–312. find
  15. Tavily. 2025. Tavily search api. tavily.com/documentation. https://docs. find
  16. Yidong Wang, Qi Guo, Wenjin Yao, Hongbo Zhang, Xin Zhang, Zhen Wu, Meishan Zhang, Xinyu Dai, Min Zhang, Qingsong Wen, Wei Ye, Shikun Zhang, and Yue Zhang. 2024. AutoSurvey: Large Language Models Can Automatically Write Surveys. Preprint, arXiv:2406.10252. find
  17. Ruibin Xiong, Yimeng Chen, Dmitrii Khizbullin, Mingchen Zhuge, and Jürgen Schmidhuber. 2025. Beyond Outlining: Heterogeneous Recursive Planning for Adaptive Long-form Writing with Language Models. Preprint, arXiv:2503.08275.
  18. OpenAI. 2025b. GPT-5. https://openai.com/ index/introducing-gpt-5/. ↗
  19. Renjun Xu and Jingwen Peng. 2025. A Comprehensive Survey of Deep Research: Systems, Methodologies, and Applications. Preprint, arXiv:2506.12594. — [arXiv:2506.12594](https://arxiv.org/abs/2506.12594) find
  20. OpenAI. 2025c. OpenAI deep research. Technical report, OpenAI. https://openai.com/index/ introducing-deep-research/. ↗
  21. Hongyang Yang, Boyu Zhang, Neng Wang, Cheng Guo, Xiaoli Zhang, Likun Lin, Junlin Wang, Tianyu Zhou, Mao Guan, Runjia Zhang, and Christina Dan Wang. 2024. FinRobot: An Open-Source AI Agent Platform for Financial Applications using Large Language Models. Preprint, arXiv:2405.14767. — [arXiv:2405.14767](https://arxiv.org/abs/2405.14767) find
  22. OpenAI. 2025d. OpenAI O1 system card. Technical report, OpenAI. https://openai.com/index/ openai-o1-system-card/. ↗
  23. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311–318.
  24. Martin Pichlmair, Riddhi Raj, and Charlene Putney. 2024. Drama Engine: A Framework for Narrative Agents. Preprint, arXiv:2408.11574. — [arXiv:2408.11574](https://arxiv.org/abs/2408.11574) find
  25. Evan F. Risko and Sam J. Gilbert. 2016. Cognitive Offloading. Trends in Cognitive Sciences, 20(9):676– 688. find
  26. Arvind Satyanarayan, Dominik Moritz, Kanit Wongsuphasawat, and Jeffrey Heer. 2017. Vega-lite: A grammar of interactive graphics. IEEE Trans. Visualization & Comp. Graphics (Proc. InfoVis). find
  27. Yijia Shao, Yucheng Jiang, Theodore A. Kanell, Peter Xu, Omar Khattab, and Monica S. Lam. 2024. Assisting in Writing Wikipedia-like Articles From Scratch with Large Language Models. Preprint, arXiv:2402.14207.
  28. Danqing Shi, Xinyue Xu, Fuling Sun, Yang Shi, and Nan Cao. 2021. Calliope: Automatic Visual Data Story Generation from a Spreadsheet. IEEE Transactions on Visualization and Computer Graphics, 27(2):453–463. find
  29. Zhaorui Yang, Bo Pan, Han Wang, Yiyao Wang, Xingyu Liu, Minfeng Zhu, Bo Zhang, and Wei Chen. 2025a. Multimodal DeepResearcher: Generating Text-Chart Interleaved Reports From Scratch with Agentic Framework. Preprint, arXiv:2506.02454. — [arXiv:2506.02454](https://arxiv.org/abs/2506.02454) find
  30. Zhongyu Yang, Jun Chen, Dannong Xu, Junjie Fei, Xiaoqian Shen, Liangbing Zhao, Chun-Mei Feng, and Mohamed Elhoseiny. 2025b. Wikiautogen: Towards multi-modal wikipedia-style article generation. Preprint, arXiv:2503.19065. — [arXiv:2503.19065](https://arxiv.org/abs/2503.19065) find
  31. Zhaoyang Yu, Jiayi Zhang, Huixue Su, Yufan Zhao, Yifan Wu, Mingyi Deng, Jinyu Xiang, Yizhang Lin, Lingxiao Tang, Yuyu Luo, Bang Liu, and Chenglin Wu. 2026. ReCode: Unify Plan and Action for Universal Granularity Control. Preprint, arXiv:2510.23564. — [arXiv:2510.23564](https://arxiv.org/abs/2510.23564) find
  32. Ran Zhang and Steffen Eger. 2024. LLM-based multi-agent poetry generation in non-cooperative environments. Preprint, arXiv:2409.03659. find
  33. Weizhi Zhang, Yangning Li, Yuanchen Bei, Junyu Luo, Guancheng Wan, Liangwei Yang, Chenxuan Xie, Yuyao Yang, Wei-Chieh Huang, Chunyu Miao, Henry Peng Zou, Xiao Luo, Yusheng Zhao, Yankai Chen, Chunkit Chan, Peilin Zhou, Xinyang Zhang, Chenwei Zhang, Jingbo Shang, and 4 others. 2025. From Web Search to… — [arXiv:2506.18959](https://arxiv.org/abs/2506.18959) find
  34. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, and 1 others. 2023. Judging llm-as-a-judge with mt-bench and chatbot arena. Advances in neural information processing systems, 36:46595–46623.
  35. Yuxiang Zheng, Dayuan Fu, Xiangkun Hu, Xiaojie Cai, Lyumanshan Ye, Pengrui Lu, and Pengfei Liu. 2025. DeepResearcher: Scaling Deep Research via Reinforcement Learning in Real-world Environments. Preprint, arXiv:2504.03160.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tian-2026-coggen-cognitively-inspired-recursive-framework/