OCCAM introduces a novel framework for cost-efficient classification that intelligently routes queries to appropriate models based on difficulty and requirements.

• Achieves up to 5x cost reduction while maintaining accuracy targets
• Uses a cascade architecture with multiple classifiers of varying capacities
• Employs intelligent routing to direct queries to the most suitable classifier
• Implements dynamic accuracy-cost management with probabilistic guarantees

For healthcare applications, OCCAM enables optimized resource allocation, allowing high-capacity models to focus on complex medical cases while simpler models handle routine classification, reducing operational costs without compromising diagnostic accuracy.

OCCAM: Towards Cost-Efficient and Accuracy-Aware Classification Inference