GenEnv: Difficulty-Aligned Co-Evolution Between LLM Agents and Environment Simulators

TL;DR

GenEnv introduces a dynamic co-evolution framework where environment simulators generate tailored tasks for LLM agents, improving performance efficiently by aligning task difficulty with the agent's capabilities.

Contribution

The paper presents GenEnv, a novel co-evolutionary framework that dynamically generates tasks to adaptively train LLM agents, reducing data requirements and enhancing scalability.

Findings

Up to +40.3% performance improvement over 7B baselines

Matches or exceeds larger model performance

Uses 3.3× less data than offline augmentation methods

Abstract

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 $\alpha$-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.