RESEARCH PAPER

WorldAgen: Unified State-Action Prediction with Test-Time World Model Training

Chi Wan; Kangrui Wang; Yuan Si; Pingyue Zhang; Manling Li

Classification

View four quadrants
Major category
WAMs
Architecture
One Model
Prediction paradigm
Joint prediction
Source review status
Not assigned

Category review. Shared Transformer heads predict actions and action-conditioned future observations; observation-based test-time training adapts the shared model from executed transitions. Action and future-observation heads share the Transformer with action-first factorization. Reading evidence

AT A GLANCE

Contribution

WorldAgen shares a Transformer between action prediction and future-observation prediction, then adapts the shared representation using exploratory target-environment transitions. Simulated manipulation improves after observation-only test-time training (TTT), while implementation ambiguities and unreported uncertainty limit the strength of the generalization claim.

Abstract

An abstract has not been added yet.

Affiliations

Northwestern University