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

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.

Training AI assistants that can use real-world tools—like booking flights, managing orders, or querying databases—requires practice environments where they can learn through trial and error. However, such environments are scarce and expensive to build by hand. We introduce ScaleEnv, a system that automatically generates diverse, realistic practice environments from scratch. ScaleEnv works by first designing virtual tools and databases, then rigorously testing that they actually work, and finally assembling them into complete, solvable tasks—much like building a flight simulator for AI agents. Agents trained in these environments show significantly better performance on entirely new, unseen tasks they have never encountered before. We also find that exposing agents to a wider variety of environments matters more than simply giving them more practice tasks within the same environments—akin to how a well-rounded education outperforms repetitive drilling. Our work provides a scalable recipe for building the practice grounds that AI agents need to become truly capable general-purpose assistants.