ScaleEnv: Scaling Environment Synthesis from Scratch for Generalist Interactive Tool-Use Agent Training

Dunwei Tu, Hongyan Hao, Hansi Yang, Yihao Chen, Yu Yang, Yueqing Sun, Xingchen Liu, Furao Shen, Qi Gu, Hui Su, Xunliang Cai
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:122403-122432, 2026.

Abstract

Training generalist agents capable of adapting to diverse scenarios requires interactive environments for self-exploration. However, interactive environments remain critically scarce, and existing synthesis methods suffer from significant limitations regarding environmental diversity and scalability. To address these challenges, we introduce ScaleEnv, a framework that constructs fully interactive environments and verifiable tasks entirely from scratch. Specifically, ScaleEnv ensures environment reliability through procedural testing, and guarantees task completeness and solvability via tool dependency graph expansion and executable action verification. By enabling agents to learn through exploration within ScaleEnv, we demonstrate significant performance improvements on unseen, multi-turn tool-use benchmarks such as $\tau^2$-Bench and VitaBench, highlighting strong generalization capabilities. Furthermore, we investigate the relationship between increasing number of domains and model generalization performance, providing empirical evidence that scaling environmental diversity is critical for robust agent learning.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-tu26a, title = {{S}cale{E}nv: Scaling Environment Synthesis from Scratch for Generalist Interactive Tool-Use Agent Training}, author = {Tu, Dunwei and Hao, Hongyan and Yang, Hansi and Chen, Yihao and Yang, Yu and Sun, Yueqing and Liu, Xingchen and Shen, Furao and Gu, Qi and Su, Hui and Cai, Xunliang}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {122403--122432}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/tu26a/tu26a.pdf}, url = {https://proceedings.mlr.press/v306/tu26a.html}, abstract = {Training generalist agents capable of adapting to diverse scenarios requires interactive environments for self-exploration. However, interactive environments remain critically scarce, and existing synthesis methods suffer from significant limitations regarding environmental diversity and scalability. To address these challenges, we introduce ScaleEnv, a framework that constructs fully interactive environments and verifiable tasks entirely from scratch. Specifically, ScaleEnv ensures environment reliability through procedural testing, and guarantees task completeness and solvability via tool dependency graph expansion and executable action verification. By enabling agents to learn through exploration within ScaleEnv, we demonstrate significant performance improvements on unseen, multi-turn tool-use benchmarks such as $\tau^2$-Bench and VitaBench, highlighting strong generalization capabilities. Furthermore, we investigate the relationship between increasing number of domains and model generalization performance, providing empirical evidence that scaling environmental diversity is critical for robust agent learning.} }
Endnote
%0 Conference Paper %T ScaleEnv: Scaling Environment Synthesis from Scratch for Generalist Interactive Tool-Use Agent Training %A Dunwei Tu %A Hongyan Hao %A Hansi Yang %A Yihao Chen %A Yu Yang %A Yueqing Sun %A Xingchen Liu %A Furao Shen %A Qi Gu %A Hui Su %A Xunliang Cai %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-tu26a %I PMLR %P 122403--122432 %U https://proceedings.mlr.press/v306/tu26a.html %V 306 %X Training generalist agents capable of adapting to diverse scenarios requires interactive environments for self-exploration. However, interactive environments remain critically scarce, and existing synthesis methods suffer from significant limitations regarding environmental diversity and scalability. To address these challenges, we introduce ScaleEnv, a framework that constructs fully interactive environments and verifiable tasks entirely from scratch. Specifically, ScaleEnv ensures environment reliability through procedural testing, and guarantees task completeness and solvability via tool dependency graph expansion and executable action verification. By enabling agents to learn through exploration within ScaleEnv, we demonstrate significant performance improvements on unseen, multi-turn tool-use benchmarks such as $\tau^2$-Bench and VitaBench, highlighting strong generalization capabilities. Furthermore, we investigate the relationship between increasing number of domains and model generalization performance, providing empirical evidence that scaling environmental diversity is critical for robust agent learning.
APA
Tu, D., Hao, H., Yang, H., Chen, Y., Yang, Y., Sun, Y., Liu, X., Shen, F., Gu, Q., Su, H. & Cai, X.. (2026). ScaleEnv: Scaling Environment Synthesis from Scratch for Generalist Interactive Tool-Use Agent Training. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:122403-122432 Available from https://proceedings.mlr.press/v306/tu26a.html.

Related Material