Beyond Simply Environment Scaling: Designing Effective Environment Distributions for Multimodal Agent Learning
Kejian Zhu, Zhuoran Jin, Dongqi Huang, Hongbang Yuan, Yupu Hao, Kang Liu, Jun Zhao
2026-08-10
Summary
Recent works train agents by constructing large-scale multimodal environment pools. However, we find that simply increasing the number of multimodal environments does not always benefit.
What's the problem?
The paper tackles this problem: Recent works train agents by constructing large-scale multimodal environment pools. However, we find that simply increasing the number of multimodal environments does not always benefit.
What's the solution?
The authors propose this solution: For diversity, we propose **Ability-aware Environment Selection (AES)** to obtain diverse environment sets. For difficulty structure, we propose **Hierarchical Difficulty Curriculum (HDC)**, which organizes curriculum learning through two difficulty levels: harness weakening and state-scale progression.
Why it matters?
Why it matters: Experiments show that AES and HDC effectively improve multimodal agent training.
Abstract
Recent works train agents by constructing large-scale multimodal environment pools. However, we find that simply increasing the number of multimodal environments does not always benefit. We further analyze the limitations in current multimodal environment distributions through a series of experiments. Based on these findings, we study how to build more effective training environment distributions from two dimensions: **diversity** and **difficulty structure**. For diversity, we propose **Ability-aware Environment Selection (AES)** to obtain diverse environment sets. For difficulty structure, we propose **Hierarchical Difficulty Curriculum (HDC)**, which organizes curriculum learning through two difficulty levels: harness weakening and state-scale progression. Experiments show that AES and HDC effectively improve multimodal agent training.