PrimeIntellect-ai/verifiers

Our library for RL environments + evals

What it solves

It provides a standardized library for creating environments used to train and evaluate Large Language Models (LLMs), specifically for reinforcement learning workflows.

How it works

The library acts as a framework for building environments that can be integrated with the Prime Intellect ecosystem, including their training framework (prime-rl) and hosted training platform. It is designed to work alongside a CLI tool for interaction and management.

Who it’s for

Developers and researchers focusing on LLM reinforcement learning who need a structured way to define and test the model's performance in specific environments.

Highlights

  • Integrated with the Environments Hub, prime-rl, and Hosted Training platform.
  • Supports the creation of environments for both training and evaluation.
  • Includes specific support for coding agents through dedicated skills and documentation.

Related

  • Project
    facebookresearch/Pearl

    A production-ready Reinforcement Learning library from Meta for building modular AI agents capable of handling complex, real-world environments with sparse feedback and dynamic action spaces.

  • Project
    NVIDIA-NeMo/Gym

    NeMo Gym is NVIDIA’s Python library for scalable, stateful evaluation and training of AI agents. It provides modular interfaces, a hub of ready‑made environments (code execution, tool‑calling, reasoning, safety, etc.), a CLI that orchestrates model, agent, and verifier servers, and hooks for RL or fine‑tuning frameworks. Designed for reproducible large‑scale benchmarking and seamless transition to training, it integrates with the broader NeMo ecosystem and supports tracing, sandboxed execution, and multi‑GPU/Slurm scaling.

  • Project
    alibaba/ROLL

    ROLL is an open‑source library for reinforcement‑learning training of large language models at scale. It provides Ray‑based distributed execution, supports many RL algorithms (PPO, GRPO, Reinforce++, etc.), works with Megatron‑Core, vLLM, SGLang, and LoRA, and offers pipelines for RL‑based fine‑tuning (RLVR) and multi‑turn agentic RL. The project includes extensive docs, hardware‑agnostic support (NVIDIA, AMD, Ascend), and built‑in observability.

  • Project
    Farama-Foundation/Gymnasium

    Gymnasium is a community‑maintained Python library that provides a standard, versioned API for reinforcement‑learning environments (classic control, Box2D, Atari, MuJoCo, etc.). It’s a fork of OpenAI’s Gym and serves as the main platform for developing and benchmarking RL algorithms.

  • Project
    rllm-org/rllm

    An open-source framework for training language agents with reinforcement learning, allowing the same agent code to be used across different backends, sandboxes, and benchmarks.