Google LLC Patent for Adversarial Bandits Policy

ChangeBridge: Patent Grants - Business Methods (G06Q)

Published March 24th, 2026

Detected March 25th, 2026

Summary

The USPTO has granted Google LLC a patent (US12585712B2) for techniques related to generating recrawl policies for commercial offer pages using adversarial bandits algorithms and machine learning. This patent covers methods for optimizing content crawling by estimating parameters like click rate and price change likelihood.

View original document View source feed page

What changed

The United States Patent and Trademark Office (USPTO) has issued patent US12585712B2 to Google LLC. This patent details "Adversarial bandits policy for crawling highly dynamic content," which involves generating multiple strategies, each treated as an arm of a K-armed adversarial bandits algorithm with reinforcement learning. The techniques also incorporate machine learning to estimate parameters such as click rate, impression rate, and likelihood of price change, aiming to improve recrawl policies for commercial offer pages.

This patent grant represents a new intellectual property for Google, specifically in the domain of AI-driven content management and optimization. While it does not impose new regulatory obligations on other entities, it signifies a technological advancement that could influence future industry practices in dynamic content crawling and advertising. Compliance officers in technology and e-commerce sectors should be aware of such patented technologies as they may impact competitive landscapes and the development of new services.

Source document (simplified)

← USPTO Patent Grants

Adversarial bandits policy for crawling highly dynamic content

Grant US12585712B2 Kind: B2 Mar 24, 2026

Assignee

GOOGLE LLC

Inventors

Michael Bendersky, Przemysław Gajda, Sergey Novikov, Marc Alexander Najork, Shuguang Han

Abstract

Techniques of generating recrawl policies for commercial offer pages include generating a multiple strategy approach using a number of different strategies. In some implementations, each strategy is an arm of a K-armed adversarial bandits algorithm with reinforcement learning. Moreover, in some implementations, the multiple strategy approach also uses a machine learning algorithm to estimate parameters such as a click rate, impression rate, and likelihood of price change, i.e., change rate, which was assumed known in the conventional approaches.