Three-in-One World Model for Marketing Intervention
A novel AI framework named the Three-in-One World Model has been introduced for marketing interventions, combining energy-based consistency, prediction, and counterfactual inference. This model employs a Deep Boltzmann Machine (DBM) to acquire fixed belief representations based on demographics, time, and previous actions and results, utilizing lightweight, task-specific adapters. It facilitates three functions: assessing energy-based consistency through DBM free energy, predicting outcomes with the help of adapters, and performing counterfactual inference by altering action inputs while keeping beliefs constant. In controlled simulations with established latent price sensitivity, promotion responsiveness, and base preferences, the adapters demonstrate comparable performance. The research can be found on arXiv under ID 2605.07199.
Key facts
- Three-in-One World Model architecture proposed
- Uses Deep Boltzmann Machine (DBM) for belief representation
- Supports energy-based consistency, prediction, and counterfactual inference
- Lightweight task-specific adapters attached to frozen DBM
- Controlled simulation with known latent consumer attributes
- Paper available on arXiv: 2605.07199
Entities
Institutions
- arXiv