ProRL Agent: Rollout-as-a-Service for RL Training of Multi-Turn LLM Agents

TL;DR

ProRL Agent introduces a scalable, API-driven infrastructure for RL training of multi-turn LLM agents, enabling flexible, maintainable, and diverse sandbox environments for complex interactive tasks.

Contribution

It presents ProRL Agent, a novel rollout-as-a-service system that decouples rollout management from training, supporting diverse tasks and environments in RL training for LLM agents.

Findings

Successfully trained LLM agents on software engineering tasks

Validated scalability and flexibility of the infrastructure

Open-sourced and integrated with NVIDIA NeMo Gym

Abstract

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.