CJReinforce/JOWA
Official code for the ICLR 2025 paper, "Scaling Offline Model-Based RL via Jointly-Optimized World-Action Model Pretraining"
This project offers a sophisticated AI agent capable of mastering numerous classic video games, outperforming human players and other advanced AI systems. It takes raw game footage (video frames) as input and produces optimal game-playing actions, enabling a single AI to excel across a wide range of distinct game environments. Game developers, AI researchers, and those interested in evaluating or advancing generalist AI for interactive systems would find this useful.
No commits in the last 6 months.
Use this if you need a pre-trained, highly capable AI agent that can quickly learn and adapt to play multiple complex video games with minimal demonstrations.
Not ideal if your primary goal is to develop simpler, task-specific AI for non-game environments or if you lack the computational resources for large-scale AI training and evaluation.
Stars
28
Forks
1
Language
Python
License
GPL-3.0
Category
Last pushed
Dec 01, 2024
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/CJReinforce/JOWA"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
galilai-group/stable-worldmodel
Reliable, minimal and scalable library for evaluating and conducting world model research
NM512/dreamerv3-torch
Implementation of Dreamer v3 in pytorch.
danijar/dreamerv2
Mastering Atari with Discrete World Models
danijar/dreamerv3
Mastering Diverse Domains through World Models
danijar/dreamer
Dream to Control: Learning Behaviors by Latent Imagination