At CraftJarvis, we're a passionate team committed to exploring the vast potential of AI in the dynamic, open-world environment of Minecraft. Our focus is on developing a generalist agent, an AI entity capable of mastering a wide range of tasks and challenges within this virtual world.
Here are a list of our latest publications on Open-world Agents. (Sort by time order)
-
Scalable Multi-Task Reinforcement Learning for Generalizable Spatial Intelligence in Visuomotor Agents (ROCKET-3)
-
JARVIS-VLA: Post-Training Large-Scale Vision Language Models to Play Visual Games with Keyboards and Mouse (ACL 2025)
-
MCU: An Evaluation Framework for Open-Ended Game Agents (ICML 2025)
-
Open-World Skill Discovery from Unsegmented Demonstrations (ICCV 2025)
-
MineStudio: A Streamlined Package for Minecraft AI Agent Development
-
ROCKET-2: Steering Visuomotor Policy via Cross-View Goal Alignment (AAAI 2026)
-
ROCKET-1: Master Open-World Interaction with Visual-Temporal Context Prompting (CVPR 2025)
-
GROOT-2: Weakly Supervised Multi-Modal Instruction Following Agents (ICLR 2025)
-
OmniJARVIS: Unified Vision-Language-Action Tokenization Enables Open-World Instruction Following Agents (NeurIPS 2024)
-
JARVIS-1: Open-world Multi-task Agents with Memory-Augmented Multimodal Language Models (T-PAMI 2024)
-
GROOT: Learning to Follow Instructions by Watching Gameplay Videos (ICLR 2024)
-
Describe, Explain, Plan and Select: Interactive Planning with Large Language Models Enables Open-World Multi-Task Agents (NeurIPS 2023)
-
Open-World Multi-Task Control Through Goal-Aware Representation Learning and Adaptive Horizon Prediction (CVPR 2023)
