SKU RLE-1401 · Sold by External
RAGEN Environments
Product specifications
| SKU | RLE-1401 |
|---|---|
| Data type | RL environments |
| Volume | 10 environments |
| Size on disk | Not published by source |
| Format | Python (Gym-compatible interface; installed from source via setup script) |
| Access model | PUBLIC LICENSE |
| Pricing | Free · open-source license |
| Quality score | — |
| License | MIT License |
RAGEN is an open-source RL framework for training reasoning agents in interactive environments, with diagnostic tooling for RL training dynamics. Its bundled StarPO/env suite ships 10 environments spanning discrete-action puzzles (Sokoban, FrozenLake, Bandit, Countdown, Sudoku) and language-based tasks (WebShop, DeepCoder, SearchQA, Lean, MetaMathQA). Environments are Gym-compatible and procedurally generated, with interactive tasks running multi-turn rollouts. Developed at Northwestern University's ML Lab (RAGEN-AI / mll-lab).