Elamur Research⌕
☰
home / papers / 2026 / CAR: Empowering Agents with Dynamic Tool Synthesis and Global Trajectory Rectification

CAR: Empowering Agents with Dynamic Tool Synthesis and Global Trajectory Rectification

Paper · 2026 · Kai Wang, Xu Wang, Zhiyuan Fu et al.

Summary

Tool-using LLM agents built on ReAct or Plan-and-Solve assume a fixed toolset. When a needed tool is missing, they fall back on long chains of atomic API calls. These chains can hit limits such as rate caps, and the agents then retry blindly. The retries fill the context with error logs and lead to hallucinated or abandoned answers. The authors propose CAR (Create And Replan), which treats reasoning as search over an action space that can grow.

CAR has four parts:

  • a planner that produces abstract sub-goals;
  • an executor that grounds each sub-goal in a concrete function call;
  • a meta-tool that writes new Python tools on demand;
  • a reflector that diagnoses failures and triggers revision of the whole remaining plan, not just local retries.

To test robustness, the authors built ToolHop-Pro from ToolHop. It removes tools and injects execution errors. CAR outperforms representative baselines in these settings, and ablations show that tool creation and replanning address different failure modes. This matters for agents deployed where APIs are incomplete or unreliable.

Key points

  • The framework is formalized as ⟨S, A_t, π, E, V⟩. The action space A_t combines standard tools, the meta-tool τ_meta, and tools synthesized so far. Execution history is stored as structured intent–observation pairs to limit context pollution.
  • ToolHop-Pro extends the ToolHop benchmark introduced by Ye et al. (2025) and tests agents under three tool conditions: Complete, Missing, and Error.
  • Ablations:
    • Without replanning, a failed first attempt at tool synthesis cannot be repaired.
    • Without tool creation, performance plateaus at about 49.96% in the Error setting.
  • Generated code runs in ephemeral Docker containers with restricted network access, adding about 1.27 s per query. Inference cost and latency are higher than for baselines.
  • Code and data are available at github.com/Zaiz-77/car. The paper mentions Qwen and GPT-4o as backbone models.

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

🎓University of Science and Technology of ChinaHefei · China108 works · 112,698 citations

 Fields, topics and concepts

📚 Shelves 4The 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 works

Cite

Cite as Kai Wang, Xu Wang, Zhiyuan Fu, Yudong Zhang and Yang Wang. 2026. CAR: Empowering Agents with Dynamic Tool Synthesis and Global Trajectory Rectification. 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.869

@inproceedings{wang2026car,
  title = {{CAR: Empowering Agents with Dynamic Tool Synthesis and Global Trajectory Rectification}},
  author = {Kai Wang, Xu Wang, Zhiyuan Fu, Yudong Zhang and Yang Wang},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.869},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.869}
}

References

28 references, 16 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Anthropic. 2024. The claude 3 model family: Opus, sonnet, haiku. Technical report, Anthropic. Model card. find
  2. Tianle Cai, Xuezhi Wang, Tengyu Ma, Xinyun Chen, and Denny Zhou. 2024. Large Language Models as Tool Makers. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenReview.net. find
  3. Shibo Hao, Yi Gu, Haodi Ma, Joshua Hong, Zhen Wang, Daisy Wang, and Zhiting Hu. 2023. Reasoning with Language Model is Planning with World Model. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 8154–8173, Singapore. Association for Computational Lingu… find
  4. Sirui Hong, Mingchen Zhuge, Jonathan Chen, Xiawu Zheng, Yuheng Cheng, Jinlin Wang, Ceyao Zhang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, Liyang Zhou, Chenyu Ran, Lingfeng Xiao, Chenglin Wu, and Jürgen Schmidhuber. 2024. MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework. In The T…
  5. 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.
  6. Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, Shashank Gupta, Bodhisattwa Prasad Majumder, Katherine Hermann, Sean Welleck, Amir Yazdanbakhsh, and Peter Clark. 2023. Self-Refine: Iterative Refinement with… find
  7. Grégoire Mialon, Roberto Dessì, Maria Lomeli, Christoforos Nalmpantis, Ramakanth Pasunuru, Roberta Raileanu, Baptiste Rozière, Timo Schick, Jane Dwivedi-Yu, Asli Celikyilmaz, Edouard Grave, Yann LeCun, and Thomas Scialom. 2023. Augmented Language Models: a Survey.
  8. Trans. Mach. Learn. Res., 2023. find
  9. OpenAI. 2023. GPT-4 Technical Report. abs/2303.08774.
  10. CoRR, Joon Sung Park, Joseph C. O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S. Bernstein. 2023. Generative Agents: Interactive Simulacra of Human Behavior. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology, UIST 2023, San Francisc… find
  11. Shishir G. Patil, Tianjun Zhang, Xin Wang, and Joseph E. Gonzalez. 2024. Gorilla: Large Language Model Connected with Massive APIs. In Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 2024, NeurIPS 2024, Vancouver, BC, Canada, December…
  12. Cheng Qian, Chi Han, Yi Fung, Yujia Qin, Zhiyuan Liu, and Heng Ji. 2023. CREATOR: Tool Creation for Disentangling Abstract and Concrete Reasoning of Large Language Models. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 6922–6939, Singapore. Association for Computatio…
  13. Yujia Qin, Shihao Liang, Yining Ye, Kunlun Zhu, Lan Yan, Yaxi Lu, Yankai Lin, Xin Cong, Xiangru Tang, Bill Qian, Sihan Zhao, Lauren Hong, Runchu Tian, Ruobing Xie, Jie Zhou, Mark Gerstein, Dahai Li, Zhiyuan Liu, and Maosong Sun. 2024. ToolLLM: Facilitating Large Language Models to Master 16000+ Real…
  14. Machel Reid, Nikolay Savinov, Denis Teplyashin, Dmitry Lepikhin, Timothy P. Lillicrap, Jean-Baptiste Alayrac, Radu Soricut, Angeliki Lazaridou, Orhan Firat, Julian Schrittwieser, Ioannis Antonoglou, Rohan Anil, Sebastian Borgeaud, Andrew M. Dai, Katie Millican, Ethan Dyer, Mia Glaese, Thibault Sotti… — Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context
  15. Timo Schick, Jane Dwivedi-Yu, Roberto Dessì, Roberta Raileanu, Maria Lomeli, Eric Hambro, Luke Zettlemoyer, Nicola Cancedda, and Thomas Scialom. 2023. Toolformer: Language Models Can Teach Themselves to Use Tools. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural I…
  16. Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. 2023. Reflexion: language agents with verbal reinforcement learning. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans,…
  17. Llama Team. 2024. The Llama 3 Herd of Models. CoRR, abs/2407.21783.
  18. Guanzhi Wang, Yuqi Xie, Yunfan Jiang, Ajay Mandlekar, Chaowei Xiao, Yuke Zhu, Linxi Fan, and Anima Anandkumar. 2024a. Voyager: An Open-Ended Embodied Agent with Large Language Models. find
  19. Trans. Mach. Learn. Res., 2024. find
  20. Lei Wang, Chen Ma, Xueyang Feng, Zeyu Zhang, Hao Yang, Jingsen Zhang, Zhiyuan Chen, Jiakai Tang, Xu Chen, Yankai Lin, Wayne Xin Zhao, Zhewei Wei, and Jirong Wen. 2024b. A survey on large language model based autonomous agents. Frontiers Comput. Sci., 18(6):186345.
  21. Lei Wang, Wanyu Xu, Yihuai Lan, Zhiqiang Hu, Yunshi Lan, Roy Ka-Wei Lee, and Ee-Peng Lim. 2023. Plan-and-Solve Prompting: Improving Zero-Shot Chain-of-Thought Reasoning by Large Language Models. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Lon… find
  22. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V. Le, and Denny Zhou. 2022. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. In Advances in Neural Information Processing Systems 35: Annual Conference on Neural Information Processi…
  23. Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, Enyu Zhou, Rui Zheng, Xiaoran Fan, Xiao Wang, Limao Xiong, Yuhao Zhou, Weiran Wang, Changhao Jiang, Yicheng Zou, Xiangyang Liu, and 9 others. 2025. The rise and potential of large language model…
  24. An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, Junyang Lin, Kai Dang, and 22 others. 2024. Qwen2.5 Technical Report. CoRR, abs/2412.1511… find
  25. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. 2023a. Tree of Thoughts: Deliberate Problem Solving with Large Language Models. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023,… find
  26. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik R. Narasimhan, and Yuan Cao. 2023b. ReAct: Synergizing Reasoning and Acting in Language Models. In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023. OpenReview.net.
  27. Junjie Ye, Zhengyin Du, Xuesong Yao, Weijian Lin, Yufei Xu, Zehui Chen, Zaiyuan Wang, Sining Zhu, Zhiheng Xi, Siyu Yuan, Tao Gui, Qi Zhang, Xuanjing Huang, and Jiecao Chen. 2025. ToolHop: A Query-Driven Benchmark for Evaluating Large Language Models in Multi-Hop Tool Use. In Proceedings of the 63rd…
  28. Andy Zhou, Kai Yan, Michal Shlapentokh-Rothman, Haohan Wang, and Yu-Xiong Wang. 2024. Language Agent Tree Search Unifies Reasoning, Acting, and Planning in Language Models. In Forty-first International Conference on Machine Learning, ICML 2024, Vienna, Austria, July 21-27, 2024. OpenReview.net. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-car-empowering-agents-dynamic-tool/