spiral-rl/spiral

Features

• Reasoning Environments - Multi-agent reinforcement learning for zero-sum games.
• Single Agent Optimization - Multi-agent reinforcement learning for incentivizing reasoning.