@inproceedings{chen-etal-2026-toolforge,
title = "{T}ool{F}orge: A Data Synthesis Pipeline for Multi-Hop, Multi-Turn, and Self-Reflective Tool-Use Data",
author = "Chen, Hao and
Zhou, Shunyuan and
Hu, Zhexin and
Chai, Jiajun and
He, Hang and
Yang, Haocheng and
Wang, Xiaohan and
Lin, Wei and
Wang, Tianyu and
Zhao, Zhuofeng and
Yin, Guojun",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1250/",
pages = "28470--28492",
ISBN = "979-8-89176-459-0",
abstract = "Training LLMs for tool use requires more than final-answer supervision: models need trajectories that specify how to decompose a question, call tools across turns, use retrieved evidence, and recover from earlier errors. Existing data construction pipelines often rely on large-scale real-world tool API calls or simple templates, making them costly or limited in interaction coverage. We introduce ToolForge, a virtual-tool data synthesis pipeline for constructing tool-use datasets with three targeted abilities: multi-hop reasoning, multi-turn tool use, and self-reflection. Starting from QA seed data, ToolForge builds an extensible tool space from 19 base virtual tools and automatically generated variants, selects among four tool-use paradigms, simulates both successful and misleading retrieval outcomes, and generates validated tool-use dialogues without large-scale real-world tool API execution. A Multi-Layer Validation module filters invalid traces and is audited with hard negatives mined by Monte Carlo Tree Search. Fine-tuning Qwen3-8B on only 4,250 ToolForge samples produces a model named ToolForge-8B, which improves over the Tool-Integrated Reasoning baseline by 21.6{\%} on five out-of-domain QA benchmarks and remains competitive on BFCL and API-Bank. We further quantify the pipeline construction cost, including tool initialization, dialogue generation, validation, and SFT, and compare ToolForge with existing synthetic tool-data sources. Our code and dataset are publicly available at https://github.com/Buycar-arb/ToolForge."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chen-etal-2026-toolforge">
<titleInfo>
<title>ToolForge: A Data Synthesis Pipeline for Multi-Hop, Multi-Turn, and Self-Reflective Tool-Use Data</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hao</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shunyuan</namePart>
<namePart type="family">Zhou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhexin</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Chai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hang</namePart>
<namePart type="family">He</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haocheng</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaohan</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wei</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tianyu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhuofeng</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guojun</namePart>
<namePart type="family">Yin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Training LLMs for tool use requires more than final-answer supervision: models need trajectories that specify how to decompose a question, call tools across turns, use retrieved evidence, and recover from earlier errors. Existing data construction pipelines often rely on large-scale real-world tool API calls or simple templates, making them costly or limited in interaction coverage. We introduce ToolForge, a virtual-tool data synthesis pipeline for constructing tool-use datasets with three targeted abilities: multi-hop reasoning, multi-turn tool use, and self-reflection. Starting from QA seed data, ToolForge builds an extensible tool space from 19 base virtual tools and automatically generated variants, selects among four tool-use paradigms, simulates both successful and misleading retrieval outcomes, and generates validated tool-use dialogues without large-scale real-world tool API execution. A Multi-Layer Validation module filters invalid traces and is audited with hard negatives mined by Monte Carlo Tree Search. Fine-tuning Qwen3-8B on only 4,250 ToolForge samples produces a model named ToolForge-8B, which improves over the Tool-Integrated Reasoning baseline by 21.6% on five out-of-domain QA benchmarks and remains competitive on BFCL and API-Bank. We further quantify the pipeline construction cost, including tool initialization, dialogue generation, validation, and SFT, and compare ToolForge with existing synthetic tool-data sources. Our code and dataset are publicly available at https://github.com/Buycar-arb/ToolForge.</abstract>
<identifier type="citekey">chen-etal-2026-toolforge</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1250/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>28470</start>
<end>28492</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T ToolForge: A Data Synthesis Pipeline for Multi-Hop, Multi-Turn, and Self-Reflective Tool-Use Data
%A Chen, Hao
%A Zhou, Shunyuan
%A Hu, Zhexin
%A Chai, Jiajun
%A He, Hang
%A Yang, Haocheng
%A Wang, Xiaohan
%A Lin, Wei
%A Wang, Tianyu
%A Zhao, Zhuofeng
%A Yin, Guojun
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F chen-etal-2026-toolforge
%X Training LLMs for tool use requires more than final-answer supervision: models need trajectories that specify how to decompose a question, call tools across turns, use retrieved evidence, and recover from earlier errors. Existing data construction pipelines often rely on large-scale real-world tool API calls or simple templates, making them costly or limited in interaction coverage. We introduce ToolForge, a virtual-tool data synthesis pipeline for constructing tool-use datasets with three targeted abilities: multi-hop reasoning, multi-turn tool use, and self-reflection. Starting from QA seed data, ToolForge builds an extensible tool space from 19 base virtual tools and automatically generated variants, selects among four tool-use paradigms, simulates both successful and misleading retrieval outcomes, and generates validated tool-use dialogues without large-scale real-world tool API execution. A Multi-Layer Validation module filters invalid traces and is audited with hard negatives mined by Monte Carlo Tree Search. Fine-tuning Qwen3-8B on only 4,250 ToolForge samples produces a model named ToolForge-8B, which improves over the Tool-Integrated Reasoning baseline by 21.6% on five out-of-domain QA benchmarks and remains competitive on BFCL and API-Bank. We further quantify the pipeline construction cost, including tool initialization, dialogue generation, validation, and SFT, and compare ToolForge with existing synthetic tool-data sources. Our code and dataset are publicly available at https://github.com/Buycar-arb/ToolForge.
%U https://aclanthology.org/2026.emnlp-main.1250/
%P 28470-28492
Markdown (Informal)
[ToolForge: A Data Synthesis Pipeline for Multi-Hop, Multi-Turn, and Self-Reflective Tool-Use Data](https://aclanthology.org/2026.emnlp-main.1250/) (Chen et al., EMNLP 2026)
ACL
- Hao Chen, Shunyuan Zhou, Zhexin Hu, Jiajun Chai, Hang He, Haocheng Yang, Xiaohan Wang, Wei Lin, Tianyu Wang, Zhuofeng Zhao, and Guojun Yin. 2026. ToolForge: A Data Synthesis Pipeline for Multi-Hop, Multi-Turn, and Self-Reflective Tool-Use Data. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 28470–28492, Budapest, Hungary. Association for Computational Linguistics.