Crafter

The 'Crafter' dataset/benchmark contains high-dimensional observation spaces used to evaluate model-based reinforcement learning agents' ability to learn abstract representations for effective planning and control.

Papers using Crafter (16)

Learning When to Plan: Efficiently Allocating Test-Time Compute for LLM Agents2025 · 9 citesDreamer-CDP: Improving Reconstruction-free World Models Via Continuous Deterministic Representation Prediction2026Subgoal Graph-Augmented Planning for LLM-Guided Open-World Reinforcement Learning2025Goal-Guided Efficient Exploration via Large Language Model in Reinforcement Learning2025HERAKLES: Hierarchical Skill Compilation for Open-ended LLM Agents2025TransDreamerV3: Implanting Transformer In DreamerV32025Causal-aware Large Language Models: Enhancing Decision-Making Through Learning, Adapting and Acting2025World Model Agents with Change-Based Intrinsic Motivation2025EDELINE: Enhancing Memory in Diffusion-based World Models via Linear-Time Sequence Modeling2025EnvGen: Generating and Adapting Environments via LLMs for Training Embodied Agents2024 · 2 citesWorld Models with Hints of Large Language Models for Goal Achieving2024Efficient World Models with Context-Aware Tokenization2024Improving Sample Efficiency of Reinforcement Learning with Background Knowledge from Large Language Models2024Instruction Following with Goal-Conditioned Reinforcement Learning in Virtual Environments2024The Interpretability of Codebooks in Model-Based Reinforcement Learning is Limited2024From Laws to Motivation: Guiding Exploration through Law-Based Reasoning and Rewards2024