Elamur Research⌕
☰
home / papers / 2026 / TInR: Exploring Tool-Internalized Reasoning in Large Language Models

TInR: Exploring Tool-Internalized Reasoning in Large Language Models

Paper · 2026 · Qiancheng Xu, Yongqi Li, Fan Liu et al.

Summary

The paper addresses a weakness of Tool-Integrated Reasoning (TIR). In TIR, LLMs call external tools while reasoning but need the tools' documentation placed in the prompt. The authors argue this causes three problems: models struggle to master heterogeneous documentation, context windows limit how many tools fit, and long prompts slow inference. They propose Tool-Internalized Reasoning (TInR), where tool knowledge is stored in the model's parameters rather than read from context. Their framework, TInR-U, assigns each tool a dedicated vocabulary token and generates calls in two steps. A <tool_token> scope first selects tools, and a <tool_call> scope then fills in parameters. Training has three phases:

  1. Tool internalization with bidirectional alignment, mapping documentation to tokens and recovering documentation from tokens.
  2. An SFT warm-up on reasoning traces built through rejection sampling.
  3. Reinforcement learning with rewards tailored to tool tokens.

In both in-domain and out-of-domain evaluations, TInR-U outperforms the baselines and is more efficient. The approach matters because it removes the need to retrieve tools or carry documentation in the prompt when the toolset is large.

Key points

  • TInR-U reports an 18.13% relative improvement in out-of-domain tool calling over prior methods.
  • On the multi-turn subsets of ToolACE and BFCL, TInR-U consistently beats ToolGen, the strongest baseline. It also outperforms the agent-style DeepAgent and a separated ToolRetriever+ToolRL pipeline.
  • The dedicated tool tokens worked better than alternative tool-identifier schemes, including hierarchical tree-based indexing.
  • In a continual-learning test, the test set was split into Task 1 (398 samples) and Task 2 (367 samples), with 50 Task 1 samples rehearsed during Task 2 training.
    • Task 1 accuracy dropped only slightly, from 60.55% to 58.29%.
    • Task 2 accuracy rose from 15.53% to about 59–60%; the paper's text and table give slightly different figures.
  • When GPT-5 was used to simulate tool updates by editing tool names and descriptions, tool-calling accuracy stayed fairly stable, moving from 61.31% to 58.25%. Code is available at github.com/travis-xu/TInR.

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 EdinburghEdinburgh · United Kingdom359 works · 77,352 citations🏛️Chinese Academy of SciencesBeijing · China194 works · 23,906 citations🎓Hong Kong Polytechnic UniversityHong Kong · China84 works · 4,956 citations🔬Shenzhen Institutes of Advanced TechnologyShenzhen · China17 works · 2,599 citations🎓Southeast UniversityNanjing · China28 works · 354 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

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

Cite

Cite as Qiancheng Xu, Yongqi Li, Fan Liu, Hongru Wang, Min Yang and Wenjie Li. 2026. TInR: Exploring Tool-Internalized Reasoning in Large Language Models. Proceedings of the 64th 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/v1/2026.acl-long.2077

@inproceedings{xu2026tinr,
  title = {{TInR: Exploring Tool-Internalized Reasoning in Large Language Models}},
  author = {Qiancheng Xu, Yongqi Li, Fan Liu, Hongru Wang, Min Yang and Wenjie Li},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.2077},
  url = {https://doi.org/10.18653/v1/2026.acl-long.2077}
}

References

38 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Guoxin Chen, Zhong Zhang, Xin Cong, Fangda Guo, Yesai Wu, Yankai Lin, Wenzheng Feng, and Yasheng Wang. 2025. Learning evolving tools for large language models. In The Thirteenth International Conference on Learning Representations. find
  2. Sijia Chen, Yibo Wang, Yi-Feng Wu, Qing-Guo Chen, Zhao Xu, Weihua Luo, Kaifu Zhang, and Lijun Zhang. 2024. Advancing tool-augmented large language models: Integrating insights from errors in inference trees. In The Thirty-eighth Annual Conference on Neural Information Processing Systems. find
  3. Tianzhe Chu, Yuexiang Zhai, Jihan Yang, Shengbang Tong, Saining Xie, Sergey Levine, and Yi Ma. 2025. SFT memorizes, RL generalizes: A comparative study of foundation model post-training. In The Second Conference on Parsimony and Learning (Recent Spotlight Track).
  4. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783.
  5. Zhibin Gou, Zhihong Shao, Yeyun Gong, yelong shen, Yujiu Yang, Minlie Huang, Nan Duan, and Weizhu Chen. 2024. ToRA: A tool-integrated reasoning agent for mathematical problem solving. In The Twelfth International Conference on Learning Representations. models with massive tools via tool embeddings.…
  6. Shadi Iskander, Sofia Tolmach, Ori Shapira, Nachshon Cohen, and Zohar Karnin. 2024. Quality matters: Evaluating synthetic data for tool-using LLMs. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 4958–4976. Association for Computational Linguistics. find
  7. Albert Q Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, et al. 2023. Mistral 7b. arXiv preprint arXiv:2310.06825. find
  8. Bowen Jin, Hansi Zeng, Zhenrui Yue, Jinsung Yoon, Sercan Arik, Dong Wang, Hamed Zamani, and Jiawei Han. 2025. Search-r1: Training llms to reason and leverage search engines with reinforcement learning.
  9. Chengpeng Li, Mingfeng Xue, Zhenru Zhang, Jiaxi Yang, Beichen Zhang, Xiang Wang, Bowen Yu, Binyuan Hui, Junyang Lin, and Dayiheng Liu. 2025a. Start: Self-taught reasoner with tools.
  10. 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.
  11. Xiaoxi Li, Wenxiang Jiao, Jiarui Jin, Guanting Dong, Jiajie Jin, Yinuo Wang, Hao Wang, Yutao Zhu, Ji-Rong Wen, Yuan Lu, and Zhicheng Dou. 2026. Deepagent: A general reasoning agent with scalable toolsets. In Proceedings of the ACM Web Conference 2026, WWW ’26, page 2219–2230. Association for Computi… find
  12. Xuefeng Li, Haoyang Zou, and Pengfei Liu. 2025c. Torl: Scaling tool-integrated rl. find
  13. Yongqi Li, Wenjie Wang, Leigang Qu, Liqiang Nie, Wenjie Li, and Tat-Seng Chua. 2024a. Generative cross-modal retrieval: Memorizing images in multimodal language models for retrieval and beyond. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long… find
  14. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948.
  15. Zekun Li, Zhiyu Chen, Mike Ross, Patrick Huber, Seungwhan Moon, Zhaojiang Lin, Xin Dong, Adithya Sagar, Xifeng Yan, and Paul Crook. 2024b. Large language models as zero-shot dialogue state tracker through function calling. In Proceedings of the 62nd Annual Meeting of the Association for Computationa… find
  16. Shibo Hao, Tianyang Liu, Zhen Wang, and Zhiting Hu. 2023. Toolkengpt: Augmenting frozen language Zhi Li, Yicheng Li, Hequan Ye, and Yin Zhang. 2024c. Towards autonomous tool utilization in language models: A unified, efficient and scalable framework. In Proceedings of the 2024 Joint International Co… find
  17. Qiqiang Lin, Muning Wen, Qiuying Peng, Guanyu Nie, Junwei Liao, Xiaoyun Mo, Jiamu Zhou, Cheng Cheng, Yin Zhao, Jun Wang, et al. 2025. Robust function-calling for on-device language model via function masking. In The Thirteenth International Conference on Learning Representations. find
  18. Weiwen Liu, Xingshan Zeng, Xu Huang, xinlong hao, Shuai Yu, Dexun Li, Shuai Wang, Weinan Gan, Zhengying Liu, Yuanqing Yu, Zezhong WANG, Yuxian Wang, Wu Ning, Yutai Hou, Bin Wang, Chuhan Wu, Wang Xinzhi, Yong Liu, Yasheng Wang, Duyu Tang, Dandan Tu, Lifeng Shang, Xin Jiang, Ruiming Tang, Defu Lian, Q… find
  19. Yanming Liu, Xinyue Peng, Jiannan Cao, Shi Bo, Yuwei Zhang, Xuhong Zhang, Sheng Cheng, Xun Wang, Jianwei Yin, and Tianyu Du. 2025b. Tool-planner: Task planning with clusters across multiple tools. In The Thirteenth International Conference on Learning Representations. find
  20. Pan Lu, Bowen Chen, Sheng Liu, Rahul Thapa, Joseph Boen, and James Zou. 2025. Octotools: An agentic framework with extensible tools for complex reasoning. In Workshop on Reasoning and Planning for Large Language Models. find
  21. Changle Qu, Sunhao Dai, Xiaochi Wei, Hengyi Cai, Shuaiqiang Wang, Dawei Yin, Jun Xu, and Ji-Rong Wen. 2025. From exploration to mastery: Enabling LLMs to master tools via self-driven interactions. In The Thirteenth International Conference on Learning Representations. find
  22. Stephen Robertson and Hugo Zaragoza. 2009. The probabilistic relevance framework: Bm25 and beyond. Foundations and Trends in Information Retrieval, 3(4):333–389. find
  23. Zhengliang Shi, Yuhan Wang, Lingyong Yan, Pengjie Ren, Shuaiqiang Wang, Dawei Yin, and Zhaochun Ren. 2025. Retrieval models aren’t tool-savvy: Benchmarking tool retrieval for large language models. In Findings of the Association for Computational Linguistics: ACL 2025, pages 24497–24524. Association… find
  24. Yunyue Su, Zhang Jinshuai, Bowen Fang, Wen Ye, Jinghao Zhang, Bowen Song, Weiqiang Wang, Qiang Liu, and Liang Wang. 2025. Toolscaler: Scalable generative tool calling via structure-aware semantic tokenization. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 556– 578.… find
  25. Qiaoyu Tang, Ziliang Deng, Hongyu Lin, Xianpei Han, Qiao Liang, and Le Sun. 2023. Toolalpaca: Generalized tool learning for language models with 3000 simulated cases. arXiv preprint arXiv:2306.05301.
  26. Hongru Wang, Cheng Qian, Wanjun Zhong, Xiusi Chen, Jiahao Qiu, Shijue Huang, Bowen Jin, Mengdi Wang, Kam-Fai Wong, and Heng Ji. 2025a. Acting less is reasoning more! teaching model to act efficiently. find
  27. Elias Lumer, Vamse Kumar Subbiah, James A. Burke, Pradeep Honaganahalli Basavaraju, and Austin Huber. 2024. Toolshed: Scale tool-equipped agents with advanced rag-tool fusion and tool knowledge bases. Preprint, arXiv:2410.14594. — [arXiv:2410.14594](https://arxiv.org/abs/2410.14594) find
  28. Renxi Wang, Xudong Han, Lei Ji, Shu Wang, Timothy Baldwin, and Haonan Li. 2025b. Toolgen: Unified tool retrieval and calling via generation. In The Thirteenth International Conference on Learning Representations. find
  29. 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
  30. Junde Wu, Jiayuan Zhu, Yuyuan Liu, Min Xu, and Yueming Jin. 2025. Agentic reasoning: A streamlined framework for enhancing LLM reasoning with agentic tools. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 28489– 28503. Associa…
  31. Cheng Qian, Emre Can Acikgoz, Qi He, Hongru WANG, Xiusi Chen, Dilek Hakkani-Tür, Gokhan Tur, and Heng Ji. 2025. ToolRL: Reward is all tool learning needs. In The Thirty-ninth Annual Conference on Neural Information Processing Systems. find
  32. 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…
  33. Qiancheng Xu, Yongqi Li, Heming Xia, and Wenjie Li. 2024. Enhancing tool retrieval with iterative feedback from large language models. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 9609–9619. Association for Computational Linguistics. find
  34. Qiancheng Xu, Yongqi Li, Heming Xia, Fan Liu, Min Yang, and Wenjie Li. 2025. PEToolLLM: Towards personalized tool learning in large language models. In Findings of the Association for Computational Linguistics: ACL 2025, pages 21488–21503. Association for Computational Linguistics. find
  35. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jing Zhou, Jin… — Qwen3 Technical Report
  36. 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
  37. Siyu Yuan, Kaitao Song, Jiangjie Chen, Xu Tan, Yongliang Shen, Kan Ren, Dongsheng Li, and Deqing Yang. 2025. EASYTOOL: Enhancing LLM-based agents with concise tool instruction. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguisti… find
  38. Jianguo Zhang, Tian Lan, Ming Zhu, Zuxin Liu, Thai Quoc Hoang, Shirley Kokane, Weiran Yao, Juntao Tan, Akshara Prabhakar, Haolin Chen, Zhiwei Liu, Yihao Feng, Tulika Manoj Awalgaonkar, Rithesh R N, Zeyuan Chen, Ran Xu, Juan Carlos Niebles, Shelby Heinecke, Huan Wang, Silvio Savarese, and Caiming Xio… — xLAM: A Family of Large Action Models to Empower AI Agent Systems

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/xu-2026-tinr-exploring-tool-internalized-reasoning/