Granted Patent B2

US 12,549,971 · App. 18/549,750 · Granted Feb 10, 2026

Inventors: Yuwei Ren (Beijing, CN); Huilin Xu (Temecula, CA)

Assignee: QUALCOMM Incorporated

H04W24/02H04L41/16

Abstract

Claims (101)

1 . A method for wireless communication at a user equipment (UE), comprising:

receiving, from a base station, a control message identifying a backbone model and at least one task-specific model, the backbone model being combinable with the at least one task-specific model to generate at least one multi-model machine learning application;

receiving, from the base station, the backbone model and a first task-specific model of the at least one task-specific model identified by the control message;

processing, by a first multi-model machine learning application that is a combination of the backbone model and the first task-specific model, one or more signals received from the base station to generate one or more outputs; and

communicating, with the base station or a wireless device, a first message based at least in part on the one or more outputs.

2 . The method of claim 1 , wherein receiving the control message comprises:

receiving the control message identifying a backbone model configuration indicating a plurality of scenarios, conditions, or tasks in which to apply the backbone model and a task-specific configuration for the first task-specific model indicating a first scenario, condition, or task of the plurality of scenarios, conditions, or tasks in which to combine the first task-specific model with the backbone model to generate the first multi-model machine learning application.

3 . The method of claim 1 , further comprising:

receiving, in the control message, at least one indication of a plurality of task-specific models and one or more scenarios, conditions, or tasks for combining a respective task-specific model with the backbone model;

receiving the plurality of task-specific models from the base station identified in the control message; and

communicating the first message based at least in part on a detected scenario, condition, or task indicating to combine the first task-specific model with the backbone model to generate the first multi-model machine learning application.

4 . The method of claim 1 , further comprising:

receiving an update control message identifying a second task-specific model that is combinable with the backbone model for generating a second multi-model machine learning application;

receiving, from the base station, the second task-specific model identified in the update control message;

processing, by the second multi-model machine learning application that is a combination of the backbone model and the second task-specific model, one or more second signals received from the base station to generate one or more second outputs; and

communicating a second message based at least in part on one or more second outputs.

5 . The method of claim 1 , further comprising:

receiving a second backbone model from the base station or a second base station and an indication that the second backbone model is combinable with a second task-specific model;

processing, by a second multi-model machine learning application that is a combination of the second backbone model and the second task-specific model, one or more second signals received from the base station or the second base station to generate one or more second outputs; and

communicating a second message based at least in part on one or more second outputs.

6 . The method of claim 5 , further comprising:

transmitting a request for the second backbone model, the second task-specific model, or both based at least in part on detecting a change in one or more conditions, one or more tasks, one or more scenarios, or any combination thereof.

7 . The method of claim 1 , wherein receiving the control message comprises:

receiving a first control message comprising a backbone model configuration for the backbone model; and

receiving a second control message comprising a task-specific configuration for the first task-specific model, wherein the one or more signals are processed by the first multi-model machine learning application in accordance with the backbone model configuration and the task-specific configuration.

8 . The method of claim 1 , wherein the control message comprises an index for the backbone model, the method further comprising:

determining that the first task-specific model is compatible with the backbone model based at least in part on the index for the backbone model.

9 . The method of claim 1 , wherein the control message indicates a task list supported by the backbone model, a scenario list supported by the backbone model, or a combination thereof, the method further comprising:

processing, by the first multi-model machine learning application, one or more second signals received from the base station to generate one or more second outputs corresponding to a task from the supported task list, corresponding to a scenario from the supported scenario list, or both.

10 . The method of claim 1 , further comprising:

transmitting a backbone model update request to the base station or a second base station based at least in part on detecting a scenario, a condition, or a task that is unsupported by the backbone model;

receiving a second backbone model and a second task-specific model based at least in part on the backbone model update request;

processing, by a second multi-model machine learning application that is a combination of the second backbone model and the second task-specific model, one or more second signals received from the base station or the second base station to generate one or more second outputs; and

communicating a second message based at least in part on one or more second outputs.

11 . An apparatus for wireless communication at a user equipment (UE), comprising:

a processor;

memory coupled with the processor; and

instructions stored in the memory and executable by the processor to cause the apparatus to:

receive, from a base station, a control message identifying a backbone model and at least one task-specific model, the backbone model being combinable with the at least one task-specific model to generate at least one multi-model machine learning application;

receive, from the base station, the backbone model and a first task-specific model of the at least one task-specific model identified by the control message;

process, by a first multi-model machine learning application that is a combination of the backbone model and the first task-specific model, one or more signals received from the base station to generate one or more outputs; and

communicate, with the base station or a wireless device, a first message based at least in part on the one or more outputs.

12 . The apparatus of claim 11 , wherein the instructions to receive the control message are executable by the processor to cause the apparatus to:

receive the control message identifying a backbone model configuration indicating a plurality of scenarios, conditions, or tasks in which to apply the backbone model and a task-specific configuration for the first task-specific model indicating a first scenario, condition, or task of the plurality of scenarios, conditions, or tasks in which to combine the first task-specific model with the backbone model to generate the first multi-model machine learning application.

13 . The apparatus of claim 11 , wherein the instructions are further executable by the processor to cause the apparatus to:

receive, in the control message, at least one indication of a plurality of task-specific models and one or more scenarios, conditions, or tasks for combining a respective task-specific model with the backbone model;

receive the plurality of task-specific models from the base station identified in the control message; and

communicate the first message based at least in part on a detected scenario, condition, or task indicating to combine the first task-specific model with the backbone model to generate the first multi-model machine learning application.

14 . The apparatus of claim 11 , wherein the instructions are further executable by the processor to cause the apparatus to:

receive an update control message identifying a second task-specific model that is combinable with the backbone model for generating a second multi-model machine learning application;

receive, from the base station, the second task-specific model identified in the update control message;

process, by the second multi-model machine learning application that is a combination of the backbone model and the second task-specific model, one or more second signals received from the base station to generate one or more second outputs; and

communicate a second message based at least in part on one or more second outputs.

15 . The apparatus of claim 11 , wherein the instructions are further executable by the processor to cause the apparatus to:

receive a second backbone model from the base station or a second base station and an indication that the second backbone model is combinable with a second task-specific model;

process, by a second multi-model machine learning application that is a combination of the second backbone model and the second task-specific model, one or more second signals received from the base station or the second base station to generate one or more second outputs; and

communicate a second message based at least in part on one or more second outputs.

16 . The apparatus of claim 15 , wherein the instructions are further executable by the processor to cause the apparatus to:

transmit a request for the second backbone model, the second task-specific model, or both based at least in part on detecting a change in one or more conditions, one or more tasks, one or more scenarios, or any combination thereof.

17 . The apparatus of claim 11 , wherein the instructions to receive the control message are executable by the processor to cause the apparatus to:

receive a first control message comprising a backbone model configuration for the backbone model; and

receive a second control message comprising a task-specific configuration for the first task-specific model, wherein the one or more signals are processed by the first multi-model machine learning application in accordance with the backbone model configuration and the task-specific configuration.

18 . The apparatus of claim 11 , wherein the control message comprises an index for the backbone model, and the instructions are further executable by the processor to cause the apparatus to:

determine that the first task-specific model is compatible with the backbone model based at least in part on the index for the backbone model.

19 . The apparatus of claim 11 , wherein the control message indicates a task list supported by the backbone model, a scenario list supported by the backbone model, or a combination thereof, and the instructions are further executable by the processor to cause the apparatus to:

process, by the first multi-model machine learning application, one or more second signals received from the base station to generate one or more second outputs corresponding to a task from the supported task list, corresponding to a scenario from the supported scenario list, or both.

20 . The apparatus of claim 11 , wherein the instructions are further executable by the processor to cause the apparatus to:

transmit a backbone model update request to the base station or a second base station based at least in part on detecting a scenario, a condition, or a task that is unsupported by the backbone model;

receive a second backbone model and a second task-specific model based at least in part on the backbone model update request;

process, by a second multi-model machine learning application that is a combination of the second backbone model and the second task-specific model, one or more second signals received from the base station or the second base station to generate one or more second outputs; and

communicate a second message based at least in part on one or more second outputs.

21 . A method for wireless communication at a base station, comprising:

transmitting, to a user equipment (UE), a control message identifying a backbone model and at least one task-specific model, the backbone model being combinable with the at least one task-specific model to generate at least one multi-model machine learning application;

transmitting, to the UE, the backbone model and a first task-specific model of the at least one task-specific model identified by the control message, and an instruction to combine to the backbone model and the first task-specific model to generate a first multi-model machine learning application; and

communicating, with the UE, a first message based at least in part on the first multi-model machine learning application.

22 . The method of claim 21 , wherein transmitting the control message comprises:

transmitting the control message identifying a backbone model configuration indicating a plurality of scenarios or tasks in which to apply the backbone model and a task-specific configuration for the first task-specific model indicating a first scenario or task of the plurality of scenarios or tasks in which to combine the first task-specific model with the backbone model to generate the first multi-model machine learning application.

23 . The method of claim 21 , further comprising:

transmitting an update control message identifying a second task-specific model that is combinable with the backbone model for generating a second multi-model machine learning application;

transmitting, to the UE, the second task-specific model identified in the update control message.

24 . The method of claim 21 , further comprising:

transmitting, in the control message, at least one indication of a plurality of task-specific models and one or more conditions for combining a respective task-specific model with the backbone model;

transmitting the plurality of task-specific models from the base station identified in the control message; and

communicating the first message based at least in part on a detected condition indicating to combine the first task-specific model with the backbone model to generate the first multi-model machine learning application.

25 . The method of claim 21 , further comprising:

transmitting a second backbone model to the UE and an indication that the second backbone model is combinable with a second task-specific model.

26 . The method of claim 21 , wherein transmitting the control message comprises:

transmitting a first control message comprising a backbone model configuration for the backbone model; and

transmitting a second control message comprising a task-specific configuration for the first task-specific model.

27 . The method of claim 21 , further comprising:

receiving a backbone model update request based at least in part on detection of a scenario, a condition, or a task that is unsupported by the backbone model;

transmitting a second backbone model and a second task-specific model based at least in part on the backbone model update request.

28 . The method of claim 21 , wherein transmitting the control message comprises:

transmitting the control message indicating a set of one or more application types supported by the backbone model, a set of one or more tasks supported by the backbone model, a set of one or more conditions for using the backbone model, one or more time periods for using the backbone model, or a combination thereof.

29 . The method of claim 21 , wherein transmitting the backbone model and the first task-specific model comprises:

transmitting an indication of a first structure for the backbone model and a first set of one or more parameters corresponding to the first structure and an indication of a second structure for the first task-specific model and a second set of one or more parameters corresponding to the second structure.

30 . An apparatus for wireless communications, comprising:

means for receiving, from a base station, a control message identifying a backbone model and at least one task-specific model, the backbone model being combinable with the at least one task-specific model to generate at least one multi-model machine learning application;

means for receiving, from the base station, the backbone model and a first task-specific model of the at least one task-specific model identified by the control message;

means for processing, by a first multi-model machine learning application that is a combination of the backbone model and the first task-specific model, one or more signals received from the base station to generate one or more outputs; and

means for communicating, with the base station or a wireless device, a first message based at least in part on the one or more outputs.

Assignments (1)

Continuity (1)

References Cited (16)