Large Tabular Models

Mastercard is investing heavily in large tabular models (LTMs) for its next-gen AI fraud detection and financial services. LTMs excel with structured data, crucial for banking. Mastercard is integrating them gradually as a complementary layer to mitigate risks of system-wide failure. The company plans to scale data input, develop APIs/SDKs for internal use, and prioritize privacy, transparency, and explainability. While LTMs promise enhanced efficiency and precision, their success depends on overcoming challenges like adversarial robustness, post-training costs, and regulatory acceptance.