Craftax

Craftax is a benchmark dataset used to evaluate goal-conditioned reinforcement learning agents by providing a diverse set of tasks and goals for the agents to learn from.

Papers using Craftax (11)

Goal-Conditioned Agents that Learn Everything All at Once2026Emergence of Exploration in Policy Gradient Reinforcement Learning via Retrying2026PACE: Parameter Change for Unsupervised Environment Design2026SCALAR: Learning and Composing Skills through LLM Guided Symbolic Planning and Deep RL Grounding2026Overcoming Valid Action Suppression in Unmasked Policy Gradient Algorithms2026CODE-SHARP: Continuous Open-ended Discovery and Evolution of Skills as Hierarchical Reward Programs2026Unsupervised Hierarchical Skill Discovery2026NiceWebRL: a Python library for human subject experiments with reinforcement learning environments2025Learning Instruction-Following Policies through Open-Ended Instruction Relabeling with Large Language Models2025Adam on Local Time: Addressing Nonstationarity in RL with Relative Adam Timesteps2024 · 3 citesSimplifying Deep Temporal Difference Learning2024 · 1 cites