NVIDIA Kumo Relational Outperforms Supervised Machine Learning on Relational Data, No Training RequiredNVIDIA Kumo Relational Outperforms Supervised Machine Learning on Relational Data, No Training Required
Pre-trained relational model predicts directly on databases without task-specific training
New model
Image from huggingface.co
NVIDIA introduced Kumo Relational, a pre-trained relational foundation model that predicts directly on enterprise databases without task-specific training, feature engineering or flattening of tables.
Given a predictive query, it gathers up to 10,000 historical examples as context and predicts outcomes such as churn, demand or fraud in a single forward pass.
NVIDIA says it is the first few-shot foundation model to outperform supervised machine learning on common relational benchmarks, beating supervised relational deep learning models and expert-engineered gradient-boosted tree pipelines across 41 tasks on 15 databases.
NVIDIA's time-series model adds retrieval and interpretability
On the SAP SALT enterprise benchmark, NVIDIA says Kumo Relational surpasses AutoGluon ensembles by 8% in mean reciprocal rank, and on RelBench V1 classification tasks it outperforms gradient-boosted trees on expert-engineered features by nearly 4 AUROC points on average.
The model scales to databases with 500B+ rows and is available through NVIDIA's open-source, GPU-native structured-data-models library or as an NVIDIA NIM.
Tasks are written in the Predictive Query Language, a declarative interface where users state what to predict rather than how, and the model is built to be called by agents.
From the source
Across 41 tasks on 15 databases, it beats supervised relational deep learning models and expert-engineered gradient-boosted tree pipelines, and on the SAP SALT enterprise benchmark it surpasses AutoGluon ensembles by 8% in mean reciprocal rank (MRR).
