公开(公告)号：US20240354346A1

公开(公告)日：2024-10-24

申请号：US18538965

申请日：2023-12-13

Applicant: QUALCOMM Incorporated

Inventor： Christopher LOTT ,  Mingu LEE ,  Wonseok JEON ,  Roland MEMISEVIC

IPC: G06F16/901 ,  G06F40/284

CPC classification number: G06F16/9027 ,  G06F40/284

Abstract: Certain aspects of the present disclosure provide techniques and apparatus for generating a response to a query input in a generative artificial intelligence model. An example method generally includes receiving a plurality of sets of tokens generated based on an input prompt and a first generative artificial intelligence model, each set of tokens in the plurality of sets of tokens corresponding to a candidate response to the input prompt; selecting, using a second generative artificial intelligence model and recursive adjustment of a target distribution associated with the received plurality of sets of tokens, a set of tokens from the plurality of sets of tokens; and outputting the selected set of tokens as a response to the input prompt.

公开(公告)号：US20240126987A1

公开(公告)日：2024-04-18

申请号：US18477515

申请日：2023-09-28

Applicant: QUALCOMM Incorporated

Inventor： Roland MEMISEVIC ,  Mingu LEE ,  Sunny Praful Kumar PANCHAL

IPC: G06F40/20

CPC classification number: G06F40/20

Abstract: A processor-implemented method includes receiving an input comprising a previous language stream, and generating an output language stream by a pre-trained language model, based on the input. The method further includes detecting a well-formed action based on patterns in the output language stream, and performing an operation, by an environment, in response to detecting the well-formed action. The operation returns a result. The method also includes appending the result to the output language stream to obtain an updated output language stream. The method includes repeating the generating, with the updated output language stream as the input, the detecting, the performing, and the appending until a termination condition is satisfied.

公开(公告)号：US20250124265A1

公开(公告)日：2025-04-17

申请号：US18544169

申请日：2023-12-18

Applicant: QUALCOMM Incorporated

Inventor： Mingu LEE ,  Kanghwan JANG ,  Christopher LOTT ,  Liang ZHANG

IPC: G06N3/048

Abstract: A processor-implemented method determines a practical domain for a following function in a following layer of an artificial neural network. The artificial neural network includes a leading function in a leading layer and the following function in the following layer, which is a subsequent consecutive layer of the artificial neural network. The method also sets a first quantization range of an output activation of the leading function based on the practical domain.

公开(公告)号：US20230376851A1

公开(公告)日：2023-11-23

申请号：US18319259

申请日：2023-05-17

Applicant: QUALCOMM Incorporated

Inventor： Mingu LEE ,  Saurabh Kedar PITRE ,  Tianyu JIANG ,  Christopher LOTT

IPC: G06N20/00

CPC classification number: G06N20/00

Abstract: Certain aspects of the present disclosure provide techniques and apparatus for performing machine learning. In one example, an input data sequence is accessed, and the input data sequence is sliced based on a slice length hyperparameter to generate a stacked slice input data representation. The stacked slice input data representation is processed with a slice attention layer to generate a stacked slice output data representation. The stacked slice output data representation is de-sliced to generate an output data sequence.

公开(公告)号：US20240428576A1

公开(公告)日：2024-12-26

申请号：US18613263

申请日：2024-03-22

Applicant: QUALCOMM Incorporated

Inventor： Tianyu JIANG ,  Manish Kumar SINGH ,  Hsin-Pai CHENG ,  Hong CAI ,  Mingu LEE ,  Kartikeya BHARDWAJ ,  Christopher LOTT ,  Fatih Murat PORIKLI

IPC: G06V10/82 ,  G06V10/70 ,  G06V10/77

Abstract: Certain aspects of the present disclosure provide techniques and apparatus for improved machine learning. A transformed version of image pixels is accessed as input to an attention layer of a machine learning model. A number of local attention operations to apply, in one transformer, to the transformed version of image pixels is selected based at least in part on a size of the transformed version of image pixels. A transformer output for the attention layer of the machine learning model is generated based on applying the number of local attention operations and at least one global attention operation to the transformed version of image pixels.

公开(公告)号：US20240386712A1

公开(公告)日：2024-11-21

申请号：US18500986

申请日：2023-11-02

Applicant: QUALCOMM Incorporated

Inventor： Apratim BHATTACHARYYA ,  Roland MEMISEVIC ,  Sunny Praful Kumar PANCHAL ,  Reza POURREZA ,  Mingu LEE ,  Pulkit MADAN

IPC: G06V10/82 ,  G06F40/10 ,  G06F40/284

Abstract: A processor-implemented method for generating grounded rationales for visual reasoning tasks includes receiving, by a first artificial neural network (ANN), an interleaved sequence of images and textual information. The first ANN extracts grid features of the images of the interleaved sequence of the images and the textual information to generate a representation of the interleaved sequence of the images and the textual information based on the grid features. A second ANN maps the grid features to a textual domain. The second ANN extracts visual information of the interleaved sequence of the images and the textual information based on the grid features in the textual domain. The second ANN determines a rationale based on the visual information. The visual information comprises one or more lower-level surrogate tasks.

公开(公告)号：US20220101827A1

公开(公告)日：2022-03-31

申请号：US17038887

申请日：2020-09-30

Applicant: QUALCOMM Incorporated

Inventor： Wonil CHANG ,  Jinseok LEE ,  Mingu LEE ,  Jinkyu LEE ,  Byeonggeun KIM ,  Dooyong SUNG ,  Jae-Won CHOI ,  Kyu Woong HWANG

IPC: G10L15/02 ,  G10L15/22

Abstract: System and method for operating an always-on ASR (automatic speech recognition) system by selecting target keywords and continuously detecting the selected target keywords in voice commands in a mobile device are provided. In the mobile device, a processor is configured to collect keyword candidates, collect usage frequency data for keywords in the keyword candidates, collect situational usage frequency data for the keywords in the keyword candidates, select target keywords from the keyword candidates based on the usage frequency data and the situational usage frequency data, and detect one or more of the target keywords in a voice command using continuous detection of the target keywords.

公开(公告)号：US20210005183A1

公开(公告)日：2021-01-07

申请号：US16920519

申请日：2020-07-03

Applicant: QUALCOMM Incorporated

Inventor： Mingu LEE ,  Jinkyu LEE ,  Hye Jin JANG ,  Kyu Woong HWANG

IPC: G10L15/16 ,  G06N3/04 ,  G06N3/08

Abstract: A method for operating a neural network includes receiving an input sequence at an encoder. The input sequence is encoded to produce a set of hidden representations. Attention-heads of the neural network calculate attention weights based on the hidden representations. A context vector is calculated for each attention-head based on the attention weights and the hidden representations. Each of the context vectors correspond to a portion of the input sequence. An inference is output based on the context vectors.

公开(公告)号：US20250021761A1

公开(公告)日：2025-01-16

申请号：US18545804

申请日：2023-12-19

Applicant: QUALCOMM Incorporated

Inventor： Arvind Vardarajan SANTHANAM ,  Joseph Binamira SORIAGA ,  Roland MEMISEVIC ,  Mingu LEE ,  Christopher LOTT

IPC: G06F40/284

Abstract: Techniques and apparatus for generating a response to a query input into a generative artificial intelligence model. An example method generally includes generating, based on an input query and a first generative artificial intelligence model, a sequence of tokens corresponding to a candidate response to the input query. The sequence of tokens and the input query are output to a second generative artificial intelligence model for verification. One or more first guidance signals for the generated sequence of tokens are received from the second generative artificial intelligence model. The candidate response to the input query is revised based on the generated sequence of tokens and the one or more first guidance signals, and the revised candidate response is output as a response to the received input query.

公开(公告)号：US20220067479A1

公开(公告)日：2022-03-03

申请号：US17274602

申请日：2019-10-08

Applicant: QUALCOMM Incorporated

Inventor： Mingu LEE ,  Wonil CHANG ,  Yeonseok KIM ,  Kyu Woong HWANG ,  Yin HUANG ,  Ruowei WANG ,  Haijun ZHAO ,  Janghoon CHO

IPC: G06N3/02 ,  G01C21/36 ,  G06N7/00

Abstract: Certain aspects of the present disclosure are generally directed to apparatus and techniques for event state detection. One example method generally includes receiving a plurality of sensor signals at a computing device, determining, at the computing device, probabilities of sub-event states based on the plurality of sensor signals using an artificial neural network for each of a plurality of time intervals, and detecting, at the computing device, the event state based on the probabilities of the sub-event states via a state sequence model.