Yandex Unveils Breakthrough AI for Tabular Data

TabM, a new neural architecture from Yandex, outperforms global competitors in analyzing spreadsheet-style data and is already being used in healthcare and business.

Yandex Research has developed a groundbreaking neural network architecture for processing tabular data. Called TabM, the system can rapidly analyze large datasets and generate highly accurate predictions—capabilities in high demand across sectors like medicine, logistics, and energy, according to a company press release.

TabM employs an ensemble approach, where multiple models analyze data independently and combine their results for improved accuracy with minimal computational overhead. In benchmarking tests across 46 datasets, TabM ranked first in performance, outperforming widely used tools like CatBoost, XGBoost, and LightGBM.

Yandex’s research in tabular data has already gained recognition from the international academic community, with citations from top universities in Germany, Singapore, South Korea, and the United States.

This advance cements Russia’s position in the global AI race and opens up new opportunities in science and industry. The integration of TabM is expected to accelerate data analysis in fields ranging from healthcare to finance, boosting predictive accuracy while lowering computing costs.