Elamur Research⌕
☰
home / papers / 2026 / SGA-MCTS: Decoupling Planning from Execution via Training-Free Atomic Experience Retrieval

SGA-MCTS: Decoupling Planning from Execution via Training-Free Atomic Experience Retrieval

Paper · 2026 · Xin Xie, Dongyun Xue, Wuguannan Yao et al.

Summary

Tool-using LLM agents need multi-step planning, but the usual options involve a trade-off. Search at inference time gives deep reasoning but is slow. Fine-tuning builds skills into weights that adapt poorly when tool schemas change. SGA-MCTS avoids both by treating planning as training-free retrieval of past experience.

  • Offline phase: Monte Carlo Tree Search, extended with meta-cognitive Plan and Reflect operators, explores tool-use tasks. A gated reward favours successful, short trajectories. Each useful step is condensed into a State-Goal-Action (SGA) atom, where concrete entities are replaced by typed slots such as <ID>.
  • Online phase: a frozen model fetches relevant atoms by combining semantic similarity with symbolic feasibility checks. It then uses them as soft hints when choosing its next action.

Across StableToolBench, ToolHop and BFCL v3, this lets an open-weights 8B model come close to GPT-5 and surpass a 32B baseline, with the largest gains on unfamiliar tools. The approach matters because the expensive reasoning is done once offline and then reused cheaply.

Key points

  • On multi-hop tool-use tasks, an 8B model in non-thinking mode reaches 44.79% success. That is 13.86 points above its zero-shot baseline, achieved without any parameter updates.
  • Abstraction compresses the experience store about 6.9×. On ToolHop, 10,685 explored actions become 1,560 atoms; StableToolBench goes from 2,388 to 213 and BFCL v3 from 1,290 to 226.
  • Compared with reasoning-heavy baselines, token use drops by roughly 2,080 tokens per task, about 76%.
  • On StableToolBench, where tool familiarity is low (S_fam ≈ 0.57), SGA reaches 43.80% versus 19.30% for LangMem with the 8B model. On BFCL v3 (S_fam ≈ 0.99), the gap is small.
  • Removing the symbolic scoring term on Qwen3-8B costs about 1.5%, mainly because the model calls tools without required arguments. Stated limitations are that results depend on the quality of offline exploration and on the diversity of the seed queries.

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

🏢Ant GroupHangzhou · China51 works · 1,211 citations

 Fields, topics and concepts

📚 Shelves 4The 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 works

🧭 Topics 1Research subjects and languages the work is about.Information retrieval & search · 552 works

Cite

Cite as Xin Xie, Dongyun Xue, Wuguannan Yao, Mingxiao Feng, Wengang Zhou, Xiang Qi, Houqiang Li and Peng Zhang. 2026. SGA-MCTS: Decoupling Planning from Execution via Training-Free Atomic Experience Retrieval. 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.60

@inproceedings{xie2026sga,
  title = {{SGA-MCTS: Decoupling Planning from Execution via Training-Free Atomic Experience Retrieval}},
  author = {Xin Xie, Dongyun Xue, Wuguannan Yao, Mingxiao Feng, Wengang Zhou, Xiang Qi, Houqiang Li and Peng Zhang},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.60},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.60}
}

References

37 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Jianlv Chen, Shitao Xiao, Peitian Zhang, Kun Luo, Defu Lian, and Zheng Liu. 2024. Bge m3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation. arXiv preprint arXiv:2402.03216.
  2. Zhixun Chen, Ming Li, Yuxuan Huang, Yali Du, Meng Fang, and Tianyi Zhou. 2025. Atlas: Agent tuning via learning critical steps. arXiv preprint arXiv:2503.02197. find
  3. Matthijs Douze, Alexandr Guzhva, Chengqi Deng, Jeff Johnson, Gergely Szilvasy, Pierre-Emmanuel Mazaré, Maria Lomeli, Lucas Hosseini, and Hervé Jégou. 2024. The faiss library.(2024). arXiv preprint arXiv:2401.08281.
  4. Lutfi Eren Erdogan, Nicholas Lee, Sehoon Kim, Suhong Moon, Hiroki Furuta, Gopala Anumanchipalli, Kurt Keutzer, and Amir Gholami. 2025. Plan-and-act: Improving planning of agents for long-horizon tasks. arXiv preprint arXiv:2503.09572. — [arXiv:2503.09572](https://arxiv.org/abs/2503.09572) find
  5. Runnan Fang, Yuan Liang, Xiaobin Wang, Jialong Wu, Shuofei Qiao, Pengjun Xie, Fei Huang, Huajun Chen, and Ningyu Zhang. 2025. Memp: Exploring agent procedural memory. arXiv preprint arXiv:2508.06433. find
  6. Jiazhan Feng, Shijue Huang, Xingwei Qu, Ge Zhang, Yujia Qin, Baoquan Zhong, Chengquan Jiang, Jinxin Chi, and Wanjun Zhong. 2025. Retool: Reinforcement learning for strategic tool use in llms. arXiv preprint arXiv:2504.11536.
  7. Zhicheng Guo, Sijie Cheng, Hao Wang, Shihao Liang, Yujia Qin, Peng Li, Zhiyuan Liu, Maosong Sun, and Yang Liu. 2024. Stabletoolbench: Towards stable large-scale benchmarking on tool learning of large language models. arXiv preprint arXiv:2403.07714. find
  8. Jiazheng Kang, Mingming Ji, Zhe Zhao, and Ting Bai. 2025. Memory os of ai agent. arXiv preprint arXiv:2506.06326.
  9. LangChain. 2025. Langmem: Long-term memory for llm agents. https://github.com/langchain-ai/ langmem. Accessed: 2025-06-01. ↗
  10. Ao Li, Yuexiang Xie, Songze Li, Fugee Tsung, Bolin Ding, and Yaliang Li. 2024. Agent-oriented planning in multi-agent systems. arXiv preprint arXiv:2410.02189. — [arXiv:2410.02189](https://arxiv.org/abs/2410.02189) find
  11. Xiaoxi Li, Wenxiang Jiao, Jiarui Jin, Guanting Dong, Jiajie Jin, Yinuo Wang, Hao Wang, Yutao Zhu, Ji-Rong Wen, Yuan Lu, et al. 2025. Deepagent: A general reasoning agent with scalable toolsets. arXiv preprint arXiv:2510.21618. — [arXiv:2510.21618](https://arxiv.org/abs/2510.21618) find
  12. Qiqiang Lin, Muning Wen, Qiuying Peng, Guanyu Nie, Junwei Liao, Jun Wang, Xiaoyun Mo, Jiamu Zhou, Cheng Cheng, Yin Zhao, et al. 2024. Hammer: Robust function-calling for on-device language models via function masking. arXiv preprint arXiv:2410.04587. — [arXiv:2410.04587](https://arxiv.org/abs/2410.04587) find
  13. Weiwen Liu, Xu Huang, Xingshan Zeng, Xinlong Hao, Shuai Yu, Dexun Li, Shuai Wang, Weinan Gan, Zhengying Liu, Yuanqing Yu, et al. 2024. Toolace: Winning the points of llm function calling. arXiv preprint arXiv:2409.00920.
  14. Tula Masterman, Sandi Besen, Mason Sawtell, and Alex Chao. 2024. The landscape of emerging ai agent architectures for reasoning, planning, and tool calling: A survey. arXiv preprint arXiv:2404.11584. find
  15. Niklas Muennighoff, Zitong Yang, Weijia Shi, Xiang Lisa Li, Li Fei-Fei, Hannaneh Hajishirzi, Luke Zettlemoyer, Percy Liang, Emmanuel Candès, and Tatsunori B Hashimoto. 2025. s1: Simple test-time scaling. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages…
  16. Siru Ouyang, Jun Yan, I Hsu, Yanfei Chen, Ke Jiang, Zifeng Wang, Rujun Han, Long T Le, Samira Daruki, Xiangru Tang, et al. 2025. Reasoningbank: Scaling agent self-evolving with reasoning memory. arXiv preprint arXiv:2509.25140. — [arXiv:2509.25140](https://arxiv.org/abs/2509.25140) find
  17. Shishir G Patil, Huanzhi Mao, Fanjia Yan, Charlie Cheng-Jie Ji, Vishnu Suresh, Ion Stoica, and Joseph E Gonzalez. 2025. The berkeley function calling leaderboard (bfcl): From tool use to agentic evaluation of large language models. In Forty-second International Conference on Machine Learning. find
  18. Yujia Qin, Shihao Liang, Yining Ye, Kunlun Zhu, Lan Yan, Yaxi Lu, Yankai Lin, Xin Cong, Xiangru Tang, Bill Qian, et al. 2023. Toolllm: Facilitating large language models to master 16000+ real-world apis. arXiv preprint arXiv:2307.16789.
  19. Changle Qu, Sunhao Dai, Xiaochi Wei, Hengyi Cai, Shuaiqiang Wang, Dawei Yin, Jun Xu, and Ji-Rong Wen. 2025. Tool learning with large language models: A survey. Frontiers of Computer Science, 19(8):198343.
  20. 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. Advances in Neural Information Processing Systems, 36:68539–68551. models. In The elev…
  21. 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…
  22. Guibin Zhang, Muxin Fu, Guancheng Wan, Miao Yu, Kun Wang, and Shuicheng Yan. 2025a. G-memory: Tracing hierarchical memory for multi-agent systems. arXiv preprint arXiv:2506.07398. — [arXiv:2506.07398](https://arxiv.org/abs/2506.07398) find
  23. Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. 2023. Reflexion: Language agents with verbal reinforcement learning. Advances in Neural Information Processing Systems, 36:8634–8652.
  24. Guibin Zhang, Haotian Ren, Chong Zhan, Zhenhong Zhou, Junhao Wang, He Zhu, Wangchunshu Zhou, and Shuicheng Yan. 2025b. Memevolve: Meta-evolution of agent memory systems. arXiv preprint arXiv:2512.18746. — [arXiv:2512.18746](https://arxiv.org/abs/2512.18746) find
  25. Charlie Snell, Jaehoon Lee, Kelvin Xu, and Aviral Kumar. 2024. Scaling llm test-time compute optimally can be more effective than scaling model parameters. arXiv preprint arXiv:2408.03314.
  26. Kai Zhang, Xiangchao Chen, Bo Liu, Tianci Xue, Zeyi Liao, Zhihan Liu, Xiyao Wang, Yuting Ning, Zhaorun Chen, Xiaohan Fu, et al. 2025c. Agent learning via early experience. arXiv preprint arXiv:2510.08558. find
  27. Zhiruo Wang, Zhoujun Cheng, Hao Zhu, Daniel Fried, and Graham Neubig. 2024. What are tools anyway? a survey from the language model perspective. arXiv preprint arXiv:2403.15452. find
  28. Georg Wölflein, Dyke Ferber, Daniel Truhn, Ognjen Arandjelovic, and Jakob Nikolas Kather. 2025. Llm agents making agent tools. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 26092–26130.
  29. BY Yan, Chaofan Li, Hongjin Qian, Shuqi Lu, and Zheng Liu. 2025. General agentic memory via deep research. arXiv preprint arXiv:2511.18423. — [arXiv:2511.18423](https://arxiv.org/abs/2511.18423) find
  30. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, et al. 2025. Qwen3 technical report. arXiv preprint arXiv:2505.09388.
  31. 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. Advances in neural information processing systems, 36:11809–11822. find
  32. Zeyu Zhang, Xiaohe Bo, Chen Ma, Rui Li, Xu Chen, Quanyu Dai, Jieming Zhu, Zhenhua Dong, and Ji-Rong Wen. 2024. A survey on the memory mechanism of large language model based agents. arXiv preprint arXiv:2404.13501. find
  33. Andrew Zhao, Daniel Huang, Quentin Xu, Matthieu Lin, Yong-Jin Liu, and Gao Huang. 2024. Expel: Llm agents are experiential learners. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 19632–19642. find
  34. Andy Zhou, Kai Yan, Michal Shlapentokh-Rothman, Haohan Wang, and Yu-Xiong Wang. 2023. Language agent tree search unifies reasoning acting and planning in language models. arXiv preprint arXiv:2310.04406. find
  35. Yuqi Zhu, Shuofei Qiao, Yixin Ou, Shumin Deng, Shiwei Lyu, Yue Shen, Lei Liang, Jinjie Gu, Huajun Chen, and Ningyu Zhang. 2025. Knowagent: Knowledge-augmented planning for llm-based agents. In Findings of the Association for Computational Linguistics: NAACL 2025, pages 3709–3732. find
  36. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. 2023b. Tree of thoughts: Deliberate problem solving with large language models. Advances in neural information processing systems, 36:11809–11822. find
  37. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik R Narasimhan, and Yuan Cao. 2022. React: Synergizing reasoning and acting in language find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/xie-2026-sga-mcts-decoupling-planning-execution/