Search papers, labs, and topics across Lattice.
This paper introduces EnvACE, a novel agentic reinforcement learning method that substitutes external environment interactions with a technique called world rehearsal, allowing agents to generate and internalize environment responses based on their actions. By alternating between acting and rehearsing, the policy learns to optimize its decision-making using task-success rewards, resulting in a robust internal world model. Experiments across multiple benchmarks demonstrate that EnvACE not only outperforms traditional environment-scaling methods but also enhances policy learning efficiency, particularly in scenarios with limited external interactions.
World rehearsal enables LLM agents to internalize environment dynamics, achieving superior performance without costly external interactions.
Training large language model agents for long-horizon tool use typically relies on interactions with real or synthesized executable environments, whose construction and verification are costly, or on external simulators that are difficult to ground. We introduce EnvACE, an agentic reinforcement learning method that replaces external environment interaction during training with world rehearsal. The policy alternates between acting and rehearsal: it first generates a tool call, then plays the role of the environment to produce the response induced by that action, and conditions subsequent decisions on the rehearsed response. Both roles are jointly optimized end-to-end using task-success rewards. Through world rehearsal, the policy internalizes the relationship between actions and their environment responses in its parameters, yielding an agent world model that directly supports decision making. Across BFCL-v4, tau^2-Bench, VitaBench, and FinMCP-Bench, EnvACE achieves strong and transferable performance, outperforming environment-scaling baselines in the overall evaluation. Controlled studies further show that world rehearsal consistently improves policy learning across model scales. At test time, the internalized world model enables private rehearsal before committed execution, yielding further gains under a moderate rehearsal budget without additional external interaction. Our findings establish world rehearsal as a new path toward scaling LLM agent training beyond the constraints of external environments. Our code is publicly available at https://github.com/Within-yao/EnvACE.
NUS