Granted Patent B1

US 12,437,238 · App. 18/908,447 · Granted Oct 7, 2025

Inventors: Shaya Zarkesh (San Francisco, CA); Lina Lukyantseva (San Francisco, CA); Rohan Bavishi (San Francisco, CA); David Luan (San Francisco, CA); John Qian (San Francisco, CA); Claire Pajot (San Francisco, CA); Fred Bertsch (San Francisco, CA); Erich Elsen (San Francisco, CA); Curtis Hawthorne (San Francisco, CA)

Assignee: Anthropic, PBC

G06N20/00G06F9/451

Abstract

Claims (33)

1. A system for generating training data to train agents to automate multimodal interface task workflows, comprising:

an intermediary interposed between an interface, comprising multimodal content, and a user, and the intermediary is configured to:

intercept one or more user-actuated actions directed towards the interface by the user, wherein the user-actuated actions, if received by the interface, execute a task on the interface;

preserve a state of the interface prior to the execution of the task, wherein the preserved state includes multimodal data comprising arbitrary-length text sequences and arbitrary-resolution images;

translate the user-actuated actions into one or more actuation commands using a transformer-based multimodal neural network comprising multi-head attention mechanisms, wherein the actuation commands are configured to trigger one or more machine-actuated actions that replicate the user-actuated actions on the interface to cause automation of the multimodal task workflow; and

generate a multimodal training dataset to train an agent to automate the task, wherein the training dataset requires the agent to process, as input, the multimodal state of the interface prior to the execution of the task, and to generate, as output, the actuation commands, wherein the translation and generation are performed using runtime interpretation logic dynamically executing on a client-side computing device.

2. The system of claim 1 , further configured to comprise an actuator, wherein the actuator is configured to receive and process the actuation commands generated by the transformer-based multimodal neural network to perform the machine-actuated actions that replicate the multimodal user-actuated actions on the interface.

3. The system of claim 1 , wherein the multimodal state of the interface prior to the execution of the task includes one or more snapshots of the arbitrary-resolution images captured from the interface that are used as direct inputs to the transformer-based multimodal neural network.

4. The system of claim 1 , wherein the multimodal state comprises detailed metadata specific to multimodal interface elements, further comprising visual layout metadata associated with arbitrary-resolution images and textual metadata linked to arbitrary-length text sequences.

5. The system of claim 1 , wherein the multimodal state includes explicit textual thoughts or annotation input by the user that provide contextualize interpretation to visual elements captured within the arbitrary-resolution images, wherein the textual thoughts are processed using multi-head attention mechanism within the transformer-based multimodal neural network.

6. The system of claim 1 , wherein the multimodal state comprises user-provided hints contextually linked to detected multimodal interface anomalies processed and interpreted through the multi-head attention mechanism s of the transformer-based multimodal neural network.

7. The system of claim 1 , wherein the multimodal state comprises a multimodal description of the task provided by the user, including combined textual descriptions and visual annotations or highlights within arbitrary-resolution images.

8. The system of claim 1 , wherein the multimodal task comprise multiple sub-tasks structured into an explicit multimodal interface workflow, each sub-task characterized by its unique multimodal state of arbitrary-resolution images and arbitrary-length textual sequences.

9. The system of claim 8 , wherein the intermediary is further configured to separately perform the interception, the preservation, the translation, and the generation for each sub-task in the plurality of sub-tasks.

10. The system of claim 8 , wherein a current multimodal sub-task in the plurality of sub-tasks is a result of executing one or more preceding sub-tasks in the plurality of sub-tasks.

11. The system of claim 10 , wherein the multimodal state of the interface prior to the execution of the current multimodal sub-task includes one or more snapshots of the interface and textual sequences corresponding to the multimodal current sub-task, one or more snapshots of the interface corresponding to the preceding sub-tasks, and one or more actuation commands corresponding to the preceding sub-tasks.

12. The system of claim 8 , wherein the multimodal interface workflow integrates distinct multimodal inputs, including arbitrary-length textual inputs and arbitrary-resolution images, simultaneously processed by the transformer-based multimodal neural network through runtime interpretation logic.

13. The system of claim 1 , wherein the user-actuated actions include clicks, hovers, scrolls, picks, text entries, and form fills.

14. The system of claim 1 , wherein the interface is part of an application.

15. The system of claim 14 , wherein the application is a web application.

16. The system of claim 14 , wherein the application is a native application.

17. The system of claim 1 , wherein the actuation commands are editable by the user.

18. The system of claim 1 , wherein the actuation commands are part of a sequence of actuation commands.

19. A computer-implemented method for generating training data to train agents to automate tasks otherwise done by users, the computer implemented method comprising:

intercepting one or more user-actuated actions directed towards an interface by a user, wherein the user-actuated actions, if received by the interface, execute a task on the interface;

preserving a state of the interface prior to the execution of the task, wherein the preserved state includes multimodal data comprising arbitrary-length text sequences and arbitrary-resolution images;

translating the user-actuated actions into one or more actuation commands using a transformer-based multimodal neural network comprising multi-head attention mechanisms, wherein the actuation commands are configured to trigger one or more machine-actuated actions that replicate the user-actuated actions on the interface to cause automation of the multimodal task workflow; and

generating a training dataset to train an agent to automate the task, wherein the training dataset requires the agent to process, as input, the state of the interface prior to the execution of the task, and to generate, as output, the actuation commands, wherein the translation and generation are performed using runtime interpretation logic dynamically executing on a client-side computing device.

20. A non-transitory computer readable storage medium impressed with computer program instructions for generating training data to train agents to automate tasks otherwise done by users, the instructions, when executed on a processor, implement a method comprising:

intercepting one or more user-actuated actions directed towards an interface by a user, wherein the user-actuated actions, if received by the interface, execute a task on the interface;

preserving a state of the interface prior to the execution of the task, wherein the preserved state includes multimodal data comprising arbitrary-length text sequences and arbitrary-resolution images;

translating the user-actuated actions into one or more actuation commands using a transformer-based multimodal neural network comprising multi-head attention mechanisms, wherein the actuation commands are configured to trigger one or more machine-actuated actions that replicate the user-actuated actions on the interface to cause automation of the multimodal task workflow; and

generating a training dataset to train an agent to automate the task, wherein the training dataset requires the agent to process, as input, the state of the interface prior to the execution of the task, and to generate, as output, the actuation commands, wherein the translation and generation are performed using runtime interpretation logic dynamically executing on a client-side computing device.

Assignments (3)

Continuity (8)

References Cited (126)

Cited By (1)

US 12,675,385