ProRL Agent: Rollout-as-a-Service for RL Training of Multi-Turn LLM Agents
This work addresses the problem of infrastructure inefficiency for researchers and developers training multi-turn LLM agents, though it is incremental as it focuses on improving existing rollout orchestration methods.
The paper tackles the challenge of generating large numbers of sandboxed rollout trajectories for RL training of multi-turn LLM agents by introducing ProRL Agent, a scalable infrastructure that serves the full agentic rollout lifecycle through an API service, resulting in a system that is open-sourced and integrated into NVIDIA NeMo Gym.
Multi-turn LLM agents are increasingly important for solving complex, interactive tasks, and reinforcement learning (RL) is a key ingredient for improving their long-horizon behavior. However, RL training requires generating large numbers of sandboxed rollout trajectories, and existing infrastructures often couple rollout orchestration with the training loop, making systems hard to migrate and maintain. Under the rollout-as-a-service philosophy, we present ProRL Agent , a scalable infrastructure that serves the full agentic rollout lifecycle through an API service. ProRL Agent also provides standardized and extensible sandbox environments that support diverse agentic tasks in rootless HPC settings. We validate ProRL Agent through RL training on software engineering, math, STEM, and coding tasks. ProRL Agent is open-sourced and integrated as part of NVIDIA NeMo Gym.