Gym-V: A Unified Vision Environment System for Agentic Vision Research
This provides a convenient foundation for training environments and evaluation toolkits to accelerate research on agentic vision-language models, addressing a domain-specific problem for researchers in vision and reinforcement learning.
The paper tackles the lack of standardized infrastructure for vision agents by introducing Gym-V, a unified platform with 179 procedurally generated visual environments across 10 domains, enabling controlled experiments and revealing that observation scaffolding is more decisive for training success than RL algorithm choice, with captions and game rules determining learning outcomes.
As agentic systems increasingly rely on reinforcement learning from verifiable rewards, standardized ``gym'' infrastructure has become essential for rapid iteration, reproducibility, and fair comparison. Vision agents lack such infrastructure, limiting systematic study of what drives their learning and where current models fall short. We introduce \textbf{Gym-V}, a unified platform of 179 procedurally generated visual environments across 10 domains with controllable difficulty, enabling controlled experiments that were previously infeasible across fragmented toolkits. Using it, we find that observation scaffolding is more decisive for training success than the choice of RL algorithm, with captions and game rules determining whether learning succeeds at all. Cross-domain transfer experiments further show that training on diverse task categories generalizes broadly while narrow training can cause negative transfer, with multi-turn interaction amplifying all of these effects. Gym-V is released as a convenient foundation for training environments and evaluation toolkits, aiming to accelerate future research on agentic VLMs.