Open Source · Nvidia · GitHub · Hugging Face
NVIDIA Kumo Tabular Sets a New Accuracy-Efficiency Frontier for Tabular Prediction
Compiled by KHAO Editorial — aggregated from 1 source. See llms.txt for citation guidance.
★ Tier-1 Source
NVIDIA Kumo Tabular, part of the NVIDIA Kumo Structured model collection, is an open foundation model for tabular data now.
Key facts
- Kumo Tabular ranks first overall with an ELO of 1950 while running 17 faster than LimiX-2 under a uniform single RTX 6000 Pro evaluation setup
- On TALENT, it achieves the top overall ranking across classification accuracy, classification log-loss, and regression RMSE, with average ranks of 6.67, 3.98, and 4.22
- In total, Kumo Tabular-Small/Medium/Large saw about 35/71/137 million artificial tables
- The second stage varies the context from 400 to 10,240 rows, and the third extends it to 60,000 rows, still with up to 100 columns
Summary
Tabular data is the backbone of enterprise machine learning. For two decades, this work has been done with gradient-boosted trees, and it has worked well. Large Language Models showed a different way of working with new tasks. Kumo Tabular is a Transformer built around the structure of a table, utilizing column, row and in-context attention as introduced in TabICL and TabPFN. In-context Learning: A final Transformer operates on the row embeddings. A head turns each query row into class probabilities for classification and 999 quantiles for regression, from which a point prediction and an uncertainty estimate follow.