Elamur Research⌕
☰
home / papers / 2025 / Agentic Reasoning: A Streamlined Framework for Enhancing LLM Reasoning with Agentic Tools

Agentic Reasoning: A Streamlined Framework for Enhancing LLM Reasoning with Agentic Tools

Paper · 2025 · Junde Wu, Jiayuan Zhu, Yuyuan Liu et al.

Summary

Reasoning LLMs trained with outcome-based reinforcement learning do well in math and code, where answers are easy to check. They struggle with open-ended, knowledge-heavy questions that need fact gathering, calculation and keeping track of many connected ideas. The authors propose Agentic Reasoning, a framework in which a reasoning model such as DeepSeek-R1 hands subtasks to auxiliary LLM agents. While reasoning, the model emits special tokens with a query. Generation pauses, the matching agent handles the request using the current context, and its answer is fed back into the chain. Experiments showed that three agents are broadly useful:

  • a Web-Search agent,
  • a Code agent,
  • a Mind-Map agent that keeps the reasoning context as a knowledge graph.

On Humanity's Last Exam, the system raises DeepSeek-R1 to a new state of the art among public models. It comes close to the proprietary OpenAI Deep Research. Domain experts also judged it helpful for automating manual research work. This suggests that open models with well-designed tools can approach closed deep-research systems. The code and data are publicly released.

Key points

  • Humanity's Last Exam: Agentic Reasoning with DeepSeek-R1 reaches 23.8% accuracy. That is 14.4 points above the base model and 2.8 points below OpenAI Deep Research (26.6%).
  • Mind-Map agent: A graph-construction LLM extracts entities and relations from the reasoning chain, following a GraphRAG-style approach. It then applies community clustering with per-cluster summaries and answers memory queries via RAG over the graph.
  • Web-Search agent: The pipeline has four stages:
    • it splits the query into search-ready sub-queries;
    • it runs them through a search engine;
    • a re-ranking model scores the results, and if the average relevance of the top 10 pages falls below a threshold, the queries are refined again;
    • RAG is applied per sub-query, and an LLM merges the answers into one snippet.
  • Werewolf test: Against seven players with over five years of experience, the model won 72% of games. Without the Mind-Map, its win rate dropped to 36%.
  • Stated limitations:
    • higher compute cost and latency, because agents are called one after another;
    • no checking of source credibility;
    • continued exposure to LLM hallucinations.

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

🎓Carnegie Mellon UniversityPittsburgh · United States539 works · 179,533 citations🎓University of OxfordOxford · United Kingdom104 works · 157,183 citations🎓National University of SingaporeSingapore · Singapore180 works · 34,304 citations🎓Mohamed bin Zayed University of Artificial IntelligenceAbu Dhabi · United Arab Emirates124 works · 5,676 citations

 Fields, topics and concepts

📚 Shelves 6The 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 worksInformation Systems · 1,437 works

🧭 Topics 2Research subjects and languages the work is about.Language models & pretraining · 3,320 worksReasoning · 1,426 works

Cite

Cite as Junde Wu, Jiayuan Zhu, Yuyuan Liu, Min Xu and Yueming Jin. 2025. Agentic Reasoning: A Streamlined Framework for Enhancing LLM Reasoning with Agentic Tools. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2025.acl-long.1383

@inproceedings{wu2025agentic,
  title = {{Agentic Reasoning: A Streamlined Framework for Enhancing LLM Reasoning with Agentic Tools}},
  author = {Junde Wu, Jiayuan Zhu, Yuyuan Liu, Min Xu and Yueming Jin},
  booktitle = {Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2025},
  doi = {10.18653/v1/2025.acl-long.1383},
  url = {https://doi.org/10.18653/v1/2025.acl-long.1383}
}

References

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

  1. Darren Edge, Ha Trinh, Newman Cheng, Joshua Bradley, Alex Chao, Apurva Mody, Steven Truitt, and Jonathan Larson. 2024. From local to global: A graph rag approach to query-focused summarization. arXiv preprint arXiv:2404.16130. find
  2. Zhen Huang, Haoyang Zou, Xuefeng Li, Yixiu Liu, Yuxiang Zheng, Ethan Chern, Shijie Xia, Yiwei Qin, Weizhe Yuan, and Pengfei Liu. 2024. O1 replication journey–part 2: Surpassing o1-preview through simple distillation, big progress or bitter lesson? arXiv preprint arXiv:2411.16489. find
  3. Shayekh Bin Islam, Md Asib Rahman, KSM Hossain, Enamul Hoque, Shafiq Joty, and Md Rizwan Parvez. 2024. Open-rag: Enhanced retrieval-augmented reasoning with open-source large language models. arXiv preprint arXiv:2410.01782. find
  4. Aaron Jaech, Adam Kalai, Adam Lerer, Adam Richardson, Ahmed El-Kishky, Aiden Low, Alec Helyar, Aleksander Madry, Alex Beutel, Alex Carney, et al. 2024. Openai o1 system card. arXiv preprint arXiv:2412.16720.
  5. M Abdul Khaliq, P Chang, M Ma, Bernhard Pflugfelder, and F Miletić. 2024. Ragar, your falsehood radar: Rag-augmented reasoning for political fact-checking using multimodal large language models. arXiv preprint arXiv:2404.12065. find
  6. Kuang-Huei Lee, Xinyun Chen, Hiroki Furuta, John Canny, and Ian Fischer. 2024. A human-inspired reading agent with gist memory of very long contexts. arXiv preprint arXiv:2402.09727.
  7. Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, et al. 2020. Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in Neural Information Processing Systems, 33:9459–9474.
  8. Aitor Lewkowycz, Anders Andreassen, David Dohan, Ethan Dyer, Henryk Michalewski, Vinay Ramasesh, Ambrose Slone, Cem Anil, Imanol Schlag, Theo Gutman-Solo, et al. 2022. Solving quantitative reasoning problems with language models. Advances in Neural Information Processing Systems, 35:3843– 3857.
  9. Xiaoxi Li, Guanting Dong, Jiajie Jin, Yuyao Zhang, Yujia Zhou, Yutao Zhu, Peitian Zhang, and Zhicheng Dou. 2025. Search-o1: Agentic search-enhanced large reasoning models. arXiv preprint arXiv:2501.05366.
  10. Aixin Liu, Bei Feng, Bing Xue, Bingxuan Wang, Bochao Wu, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, et al. 2024. Deepseek-v3 technical report. arXiv preprint arXiv:2412.19437. find
  11. Xusheng Luo, Shaojun Xu, and Changliu Liu. 2023. Obtaining hierarchy from human instructions: an llms-based approach. In CoRL 2023 Workshop on Learning Effective Abstractions for Planning (LEAP). find
  12. Grégoire Mialon, Clémentine Fourrier, Craig Swift, Thomas Wolf, Yann LeCun, and Thomas Scialom. 2023. Gaia: a benchmark for general ai assistants. arXiv preprint arXiv:2311.12983.
  13. OpenAI. Learning to reason with LLMs. Charles Packer, Vivian Fang, Shishir_G Patil, Kevin Lin, Sarah Wooders, and Joseph_E Gonzalez. 2023. Memgpt: Towards llms as operating systems. find
  14. Keivalya Pandya and Mehfuza Holia. 2023. Automating customer service using langchain: Building custom open-source gpt chatbot for organizations. arXiv preprint arXiv:2310.05421. — [arXiv:2310.05421](https://arxiv.org/abs/2310.05421) find
  15. Long Phan, Alice Gatti, Ziwen Han, Nathaniel Li, Josephina Hu, Hugh Zhang, Chen Bo Calvin Zhang, Mohamed Shaaban, John Ling, Sean Shi, Michael Choi, Anish Agrawal, Arnav Chopra, Adam Khoja, Ryan Kim, Richard Ren, Jason Hausenloy, Oliver Zhang, Mantas Mazeika, Tung Nguyen, Daron Anderson, Imad Ali Sh… — Humanity's Last Exam
  16. Di Zhang, Jianbo Wu, Jingdi Lei, Tong Che, Jiatong Li, Tong Xie, Xiaoshui Huang, Shufei Zhang, Marco Pavone, Yuqiang Li, et al. 2024. Llama-berry: Pairwise optimization for o1-like olympiad-level mathematical reasoning. arXiv preprint arXiv:2410.02884. find
  17. Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, and Yanlin Wang. 2024. Memorybank: Enhancing large language models with long-term memory. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 19724–19731. find
  18. Yiwei Qin, Xuefeng Li, Haoyang Zou, Yixiu Liu, Shijie Xia, Zhen Huang, Yixin Ye, Weizhe Yuan, Hector Liu, Yuanzhi Li, et al. 2024. O1 replication journey: A strategic progress report–part 1. arXiv preprint arXiv:2410.18982. find
  19. David Rein, Betty Li Hou, Asa Cooper Stickland, Jackson Petty, Richard Yuanzhe Pang, Julien Dirani, Julian Michael, and Samuel R Bowman. 2023. Gpqa: A graduate-level google-proof q&a benchmark. arXiv preprint arXiv:2311.12022.
  20. 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. arXiv preprint arXiv:2402.14207.
  21. DeepSeek Team. 2024. Deepseek-r1-lite-preview is now live: unleashing supercharged reasoning power. find
  22. Qwen Team. Qwq: Reflect deeply on the boundaries of the unknown, november 2024. URL https://qwenlm.github. io/blog/qwq-32b-preview. ↗
  23. Vincent A Traag, Ludo Waltman, and Nees Jan Van Eck. 2019. From louvain to leiden: guaranteeing well-connected communities. Scientific reports, 9(1):1– 12. find
  24. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824–24837.
  25. Junde Wu, Jiayuan Zhu, Yunli Qi, Jingkun Chen, Min Xu, Filippo Menolascina, and Vicente Grau. 2024. Medical graph rag: Towards safe medical large language model via graph retrieval-augmented generation. arXiv preprint arXiv:2408.04187. — [arXiv:2408.04187](https://arxiv.org/abs/2408.04187) find
  26. An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, et al. 2024. Qwen2. 5 technical report. arXiv preprint arXiv:2412.15115. find
  27. Hui Yang, Sifu Yue, and Yunzhong He. 2023. Auto-gpt for online decision making: Benchmarks and additional opinions. arXiv preprint arXiv:2306.02224. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wu-2025-agentic-reasoning-streamlined-framework-enhancing/