Pandora: Towards General World Model with Natural Language Actions and Video States
World models simulate future states of the world in response to different actions. They facilitate interactive content creation and provides a foundation for grounded, long-horizon reasoning. Current foundation models do not fully meet the capabilities of general world models: large language models...
Gespeichert in:
Hauptverfasser: | , , , , , , , , , , , , |
---|---|
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | World models simulate future states of the world in response to different
actions. They facilitate interactive content creation and provides a foundation
for grounded, long-horizon reasoning. Current foundation models do not fully
meet the capabilities of general world models: large language models (LLMs) are
constrained by their reliance on language modality and their limited
understanding of the physical world, while video models lack interactive action
control over the world simulations. This paper makes a step towards building a
general world model by introducing Pandora, a hybrid autoregressive-diffusion
model that simulates world states by generating videos and allows real-time
control with free-text actions. Pandora achieves domain generality, video
consistency, and controllability through large-scale pretraining and
instruction tuning. Crucially, Pandora bypasses the cost of
training-from-scratch by integrating a pretrained LLM (7B) and a pretrained
video model, requiring only additional lightweight finetuning. We illustrate
extensive outputs by Pandora across diverse domains (indoor/outdoor,
natural/urban, human/robot, 2D/3D, etc.). The results indicate great potential
of building stronger general world models with larger-scale training. |
---|---|
DOI: | 10.48550/arxiv.2406.09455 |