DRIVE
Implementations accompanying research on DRIVE, a decentralized peer-incentivization mechanism for emergent cooperation under changing rewards that exchanges local reward differences so incentives adapt automatically to shifted, scaled, or perturbed rewards.
REACT
Implementations and video demonstrations acompanying research on revealing evolutionary action consequence trajectories (REACT) for interpretable reinforcement learning.
Quantum Circuit Designer
A gymnasium-based set of environments for benchmarking reinforcement learning for state preparation and unitary composition in quantum circuit design.
Niryo Gymnasium
A reinforcement learning MuJoCo gymnasium environment for benchmarking robotic grasping using the Niryo-NED2 robot arm.
Emergence in Multi-Agent Systems
Implementations accompanying research on a safety persepective on emergence in multi-agent systems and mitigation techniques.
DIRECT
Implementations accompanying research on discriminative reward co-training (DIRECT), a self-imitation architecture for robust reinforcement learning from sparse rewards.
hyphi-gym
A Gymnasium benchmark suite for evaluating the robustness and multi-task performance of reinforcement learning algorithms in various discrete and continuous environments.
CROP
Implementations accompanying research on distributional-shift robust reinforcement learning using compact reshaped observation processing (CROP).
Karting Challenge
Unity project for a karting-based ML‑Agents challenge. Drive a kart around a modular track, collect rewards, and train agents to complete laps robustly across different track parts and obstacles.