JANUS: Structured Bidirectional Generation for Guaranteed Constraints and Analytical Uncertainty

Researchers introduce JANUS, a new AI framework that solves the 'Quadrilemma' in synthetic data generation by achieving high fidelity, logical constraint control, reliable uncertainty estimation, and computational efficiency simultaneously. The system uses Bayesian Decision Trees and a novel Reverse-Topological Back-filling algorithm to guarantee 100% constraint satisfaction while being 128x faster than existing methods.

• →JANUS addresses four critical challenges in synthetic data generation that existing models struggle to solve simultaneously.
• →The framework achieves 100% constraint satisfaction without rejection sampling through innovative Reverse-Topological Back-filling.
• →Analytical Uncertainty Decomposition provides 128x faster uncertainty estimation compared to Monte Carlo methods.
• →Testing across 15 datasets and 523 scenarios shows state-of-the-art fidelity with Detection Score of 0.497.
• →The system eliminates mode collapse on imbalanced data and handles complex inter-column constraints where other methods fail.