METHOD FOR GENERATING TRAINING DATA, AND ELECTRONIC DEVICE

ChangeBridge: Patent Apps - AI & Computing (G06N)

Published March 26th, 2026

Detected April 1st, 2026

Summary

USPTO published patent application US20260087385A1 filed by Beijing Baidu Netcom Science Technology Co., Ltd. on March 26, 2026. The patent covers methods for generating AI training data using state transition images derived from user interaction data, trajectory data generation, and multimodal model-based reasoning reference extraction for training interaction agent models. The inventors are Le Zhang, Yu Shi, and Jingbo Zhou.

View original document View source feed page

What changed

Baidu has been granted a US patent publication (Application No. 19403504, filed November 28, 2025) covering a method for generating AI training data. The method involves obtaining application interaction data, generating state transition images, producing trajectory data, extracting reasoning reference information via a multimodal model, and synthesizing training data for interaction agent models. This covers the full pipeline for creating training datasets from user interaction patterns.

This publication is informational for IP and compliance tracking purposes. Technology companies developing AI training pipelines or interaction agents should monitor this patent for potential licensing implications or design-around considerations. No immediate compliance action is required as patent applications do not create binding obligations until granted and enforced.

Source document (simplified)

← USPTO Patent Applications

METHOD FOR GENERATING TRAINING DATA, AND ELECTRONIC DEVICE

Application US20260087385A1 Kind: A1 Mar 26, 2026

Assignee

BEIJING BAIDU NETCOM SCIENCE TECHNOLOGY CO., LTD.

Inventors

Le Zhang, Yu Shi, Jingbo Zhou

Abstract

A method for generating training data is performed by an electronic device, including: obtaining interaction data with an application, determining a state transition image based on the interaction data, generating a plurality of pieces of trajectory data based on the state transition image, obtaining reasoning reference information corresponding to the operation data in the trajectory data output by a multimodal model, by inputting the trajectory data into the multimodal model, and generating training data for training an interaction agent model based on the trajectory data and the reasoning reference information corresponding to the operation data in the trajectory data.