ViBR-WM: Visual Bayesian Regression for World Modeling

2026-09-27

AI Summary

This study addresses the challenges of temporal dependency and uncertainty prediction in world models by proposing a multimodal state modeling framework that integrates visual features, physical history, and covariates. Methodologically, it employs a modular architecture to achieve interpretable regression of trend, seasonal, and periodic dynamics. Bayesian variable selection is introduced for dimensionality reduction, while posterior probability-weighted ensemble forecasting effectively handles parameter uncertainty and future perturbations. Furthermore, visual compression and recurrent joint state prediction techniques are incorporated to enhance modeling accuracy. Evaluated on object motion, vegetation greenness, and solar energy forecasting tasks, the proposed framework achieves significantly lower errors on physical targets compared to mainstream baselines such as ConvLSTM.

Abstract

Modeling temporal dependence and uncertainty is central to forecasting with world models. The Visual Bayesian Regression World Model combines visual features, physical histories and known covariates through interpretable regression, within a modular architecture supporting trend, seasonal and cycle dynamics. Visual compression reduces representation dimension, while Bayesian variable selection reduces active regression dimension. Posterior prediction combines forecasts across predictor subsets using their posterior probabilities as weights and accounts for parameter uncertainty and future disturbances. The model forecasts joint visual--physical states recursively and physical targets directly. Across four forecasting tasks spanning object motion, vegetation greenness and solar power, ViBR-WM achieves lower mean overall physical-target error than Temporal Straightening, ConvLSTM, PredRNN and SimVP on every task. Repeated fitting and resampling support these overall gains.

Problem

• world models
• temporal dependence
• uncertainty
• forecasting

Innovation

• Visual Bayesian Regression
• World Model
• Bayesian Variable Selection
• Visual Compression
• Posterior Prediction