--- title: "ToolGrad — paper by Zhou et al. (2026) — summary and ideas" description: "It first builds a verified tool-use workflow and only afterwards writes a matching user query and response in one LLM step. 1 reference." authors: ["Zhongyi Zhou", "Kohei Uehara", "Haoyu Zhang", "Jingtao Zhou", "Lin Gu", "Ruofei Du", "Zheng Xu", "Tatsuya Harada"] year: 2026 venue: "Findings of the Association for Computational Linguistics: ACL 2026" source: "https://aclanthology.org/2026.findings-acl.950/" doi: "10.18653/v1/2026.findings-acl.950" licence: "CC BY 4.0 (ACL Anthology)" licence_url: "https://creativecommons.org/licenses/by/4.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/zhou-2026-toolgrad-efficient-tool-use-dataset/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/zhou-2026-toolgrad-efficient-tool-use-dataset/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/53/) / ToolGrad: Efficient Tool-use Dataset Generation with Textual “Gradients” # ToolGrad: Efficient Tool-use Dataset Generation with Textual “Gradients” **Authors:** [Zhongyi Zhou](https://md.research.elamur.ai/free/en/authors/zhongyi-zhou/), Kohei Uehara, [Haoyu Zhang](https://md.research.elamur.ai/free/en/authors/haoyu-zhang/), Jingtao Zhou, Lin Gu, Ruofei Du, Zheng Xu and Tatsuya Harada · **Year:** 2026 · **Venue:** [Findings of the Association for Computational Linguistics: ACL 2026](https://md.research.elamur.ai/free/en/venues/findings/) ([Findings of the Association for Computational Linguistics 2026](https://md.research.elamur.ai/free/en/venues/findings/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/) **Links:** [Publisher's page](https://aclanthology.org/2026.findings-acl.950/) · [PDF](https://aclanthology.org/2026.findings-acl.950.pdf) · [DOI 10.18653/v1/2026.findings-acl.950](https://doi.org/10.18653/v1/2026.findings-acl.950) · [OpenAlex](https://openalex.org/W7166856381) **Citations:** 0 citations (OpenAlex, 2026-10-14) ## Summary Training LLMs to call tools requires pairs of user requests and correct tool-call sequences. Earlier synthetic pipelines such as ToolBench start from an invented query and then search for a solution with DFS. This search is costly and often fails, and it can leave unsolvable queries or flawed traces in the data. ToolGrad flips the order. It first builds a verified tool-use workflow and only afterwards writes a matching user query and response in one LLM step. The workflow is grown iteratively over mini-batches of APIs drawn from a large database. Four agent modules propose, execute and select APIs, then update the workflow. LLM feedback acts as textual "gradients", an idea borrowed from TextGrad. Using Gemini-2.5-flash-lite as the teacher, the authors built ToolGrad-500. It contains more complex tool use than ToolBench and was produced more cheaply with a near-perfect pass rate. Gemma-3 models fine-tuned on it beat ToolBench-trained baselines and several proprietary models, including on unseen tools. This suggests answer-first generation lets students surpass their teacher. Code, data and models are released. ## Key points - **Dataset and pass rate:** ToolGrad-500 holds about 500 triplets of query, API workflow and response. Every sample starts from a valid tool call, and failed steps never enter the data, which gives an almost 100% pass rate. - **BFCL results:** ToolGrad-12B places second on BFCL, only 0.1 behind Gemini 2.5 Pro. It outperforms ToolACE and its own teacher, Gemini-2.5-flash-lite. The largest gains appear on the "Multiple Parallel" subtask. - **Student beats teacher:** even the 1B fine-tuned model surpasses its teacher model on the paper's benchmarks. - **Iteration scaling:** pass rate levels off between 8 and 12 iterations, so 10 iterations were used to build ToolGrad-500. - **Sample scaling:** Gemma-3-4B was trained on 100 to 2k samples. Accuracy rises and then falls as data grows, because generated tool uses become repetitive. The authors attribute this to the lack of a shared memory module. *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 - [The University of Tokyo](https://md.research.elamur.ai/free/en/universities/the-university-of-tokyo/) — University, Tokyo, Japan · 96 works · 9,600 citations - [Tohoku University](https://md.research.elamur.ai/free/en/universities/tohoku-university/) — University, Sendai, Japan · 23 works · 595 citations - [RIKEN Center for Advanced Intelligence Project](https://md.research.elamur.ai/free/en/labs/riken-center-for-advanced-intelligence-project/) — Research institute, Tokyo, Japan · 16 works · 957 citations - [RIKEN](https://md.research.elamur.ai/free/en/labs/riken/) — Research institute, Wako, Japan · 30 works · 607 citations 📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works · [🗄️ Information Systems](https://md.research.elamur.ai/free/en/shelves/computer-science/information-systems/) · 1,730 works ## Cite Cite as: Zhongyi Zhou, Kohei Uehara, Haoyu Zhang, Jingtao Zhou, Lin Gu, Ruofei Du, Zheng Xu and Tatsuya Harada. 2026. ToolGrad: Efficient Tool-use Dataset Generation with Textual “Gradients”. Findings of the Association for Computational Linguistics: ACL 2026. The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{zhou2026toolgrad, title = {{ToolGrad: Efficient Tool-use Dataset Generation with Textual “Gradients”}}, author = {Zhongyi Zhou, Kohei Uehara, Haoyu Zhang, Jingtao Zhou, Lin Gu, Ruofei Du, Zheng Xu and Tatsuya Harada}, booktitle = {Findings of the Association for Computational Linguistics: ACL 2026}, year = {2026}, doi = {10.18653/v1/2026.findings-acl.950}, url = {https://doi.org/10.18653/v1/2026.findings-acl.950} } ``` - [← previous ToolCPT: Improving Tool Utilization in LLM Agents via Continuous Pre-training](https://md.research.elamur.ai/free/en/papers/yang-2026-toolcpt-improving-tool-utilization-llm/) - [next → Tools for Estimating the Perceived Level of Phonetic Reduction](https://md.research.elamur.ai/free/en/papers/ward-2026-tools-estimating-perceived-level-phonetic/) - [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/53/) ## References 1 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. Xiaohe Bo, Zeyu Zhang, Quanyu Dai, Xueyang Feng, Lei Wang, Rui Li, Xu Chen, and Ji-Rong Wen. 2024. Reflective multi-agent collaboration based on large language models. Advances in Neural Information Processing Systems, 37:138595–138631. \#\# Limitations [find](https://scholar.google.com/scholar?q=Xiaohe+Bo%2C+Zeyu+Zhang%2C+Quanyu+Dai%2C+Xueyang+Feng%2C+Lei+Wang%2C+Rui+Li%2C+Xu+Chen%2C+and+Ji%2DRong+Wen%2E+2024%2E+Reflective+multi%2Dagent+collaboration+based+on+large+language+models%2E+Advances+in+Neural+Information+P) *Updated 2 November 2026*