GenEnv: Difficulty-Aligned Co-Evolution Between LLM Agents and Environment Simulators
This addresses the data bottleneck in training LLM agents, offering a more efficient and adaptive approach, though it appears incremental as it builds on existing curriculum learning and simulation ideas.
The paper tackles the high cost and static nature of real-world interaction data for training LLM agents by introducing GenEnv, a framework that co-evolves agents with a generative environment simulator, resulting in up to 40.3% performance improvement over baselines and better data efficiency compared to offline methods.
Training capable Large Language Model (LLM) agents is critically bottlenecked by the high cost and static nature of real-world interaction data. We address this by introducing GenEnv, a framework that establishes a difficulty-aligned co-evolutionary game between an agent and a scalable, generative environment simulator. Unlike traditional methods that evolve models on static datasets, GenEnv instantiates a dataevolving: the simulator acts as a dynamic curriculum policy, continuously generating tasks specifically tailored to the agent's ``zone of proximal development''. This process is guided by a simple but effective $α$-Curriculum Reward, which aligns task difficulty with the agent's current capabilities. We evaluate GenEnv on five benchmarks, including API-Bank, ALFWorld, BFCL, Bamboogle, and TravelPlanner. Across these tasks, GenEnv improves agent performance by up to \textbf{+40.3\%} over 7B baselines and matches or exceeds the average performance of larger models. Compared to Gemini 2.5 Pro-based offline data augmentation, GenEnv achieves better performance while using 3.3$\times$ less data. By shifting from static supervision to adaptive simulation, GenEnv provides a data-efficient pathway for scaling agent capabilities.