Training Machine-Learned Models with Temporal Conditioning for Time-Aware Inference

ChangeBridge: Patent Apps - AI & Computing (G06N)

Published March 26th, 2026

Detected March 31st, 2026

Summary

USPTO published patent application US20260087404A1 for a method of training machine-learned models with temporal conditioning to enable time-aware inference. The application, filed by Florian Nils Hartmann and Matthew Sharifi, covers extracting temporal features from source data to construct training inputs and generate content predictions using computed loss functions. CPC classification is G06N 20/00 (Machine Learning).

View original document View source feed page

What changed

USPTO published patent application US20260087404A1 titled 'Training Machine-Learned Models with Temporal Conditioning for Time-Aware Inference.' The application discloses a method comprising: extracting temporal feature values from source data items using a temporal feature extraction system; constructing training inputs containing content from source data and extracted temporal features; generating content predictions using a machine-learned model; and computing content prediction loss to train the model. The patent covers technology for time-aware inference in machine learning systems.

This is a routine patent publication with no compliance requirements or deadlines for external parties. Technology companies developing machine learning systems may wish to review this application to understand the scope of claimed innovations and assess potential licensing needs or freedom-to-operate implications. No regulatory action is required.

Source document (simplified)

← USPTO Patent Applications

Training Machine-Learned Models with Temporal Conditioning for Time-Aware Inference

Application US20260087404A1 Kind: A1 Mar 26, 2026

Inventors

Florian Nils Hartmann, Matthew Sharifi

Abstract

An example method includes processing, using a temporal feature extraction system, a source data item to extract a temporal feature value associated with the source data item. The example method includes constructing a respective training input for a respective training example, the respective training input. The example method includes content obtained from the source data item. The example method includes the extracted temporal feature value. The example method includes generating, using a machine-learned model, a respective training output based on the respective training input, wherein the respective training output includes a content prediction. The example method includes computing, using the respective training output and a respective content evaluation signal for the respective training example, a content prediction loss. The example method includes training the machine-learned model using the content prediction loss.