EnvCraft: Synthesizing Executable Environments in Agentic RL for Claw-like Agent

Researchers introduced EnvCraft, an automated framework that synthesizes executable environments and training data for Agentic Reinforcement Learning, detailed in arXiv paper 2609.05576v1. EnvCraft's environment synthesis engine builds sandbox-isolated workspaces and its topology-aware data generation engine produces coherent task trajectories, yielding 139 interactive environments with approximately 20,000 complex tasks. Experiments on Qwen3/3.5 models ranging from 8B to 32B parameters showed gains of up to +11.9% on Claw-style benchmarks and +8.0% on general tool-use benchmarks, with concurrent reductions in inference token cost.

arXiv:2609.05576v1 Announce Type: new Abstract: The paradigm of LLMs has rapidly shifted from passive language interfaces to autonomous Claw-like agents that execute long-horizon tasks across stateful workspaces. While Agentic Reinforcement Learning Agentic RL provides a promising path to optimize these agents, its scaling is heavily bottlenecked by the severe scarcity of interactive training environments. Existing synthetic environments are strictly limited to tool-calling endpoints, rendering them insufficient for accommodating the end-to-end real-world demands of claw-like agents. To bridge this gap, we introduce EnvCraft, an automated framework for synthesizing executable environments and scalable training data. Specifically, EnvCraft employs an environment synthesis engine to build sandbox-isolated workspaces, alongside a topology-aware data generation engine to produce coherent task trajectories. Overall, we synthesize 139 interactive environments comprising approximately 20K complex tasks for Agentic RL training. Experiments on Qwen3/3.5 models 8B-32B show that our method yields gains of up to +11.9% on Claw-style benchmarks and +8.0% on general tool-use benchmarks, with concurrent reductions in inference token cost. The results confirm that synthesized executable environments provide robust and generalizable learning signals for training.