Patents by Inventor Hao Kang

Hao Kang has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).

  • Publication number: 20260350372
    Abstract: A data processing system implements analyzing a textual input using one or more layers of an LLM to obtain intermediate results; providing the intermediate results as an input to an attention mechanism of the LLM; analyzing the intermediate results using the attention mechanism to generate attention scores; providing the attention scores as an input to a softmax approximation framework that performs an exponentiation operation and normalization operation on the attention scores to generate a probability distribution; determining an integer portion of the exponentiation operation using the softmax approximation framework using a lookup table to determine a value for the integer portion of the exponentiation operation; determining a decimal portion of the exponentiation operation by performing a polynomial approximation to determine a value for the decimal portion of the exponentiation operation; and outputting the probability distribution from the attention mechanism to one or more additional layers of the lar
    Type: Application
    Filed: April 30, 2025
    Publication date: November 5, 2026
    Applicant: Microsoft Technology Licensing, LLC
    Inventors: Vedula Venkata Srikant BHARADWAJ, Saravanakumar RAJMOHAN, Victor Jonas RÜHLE, Hao KANG
  • Publication number: 20260348222
    Abstract: The present application proposes a method for controlling a vehicle partition glass based on an SOA architecture, a device, a medium and a vehicle. The SOA architecture at least includes a scenario service layer, a system service layer, an enhanced service layer, an atomic service layer, an I/O abstraction layer, and a basic service layer. The method includes: receiving, by a partition-glass control enhanced service of the enhanced service layer, at least one of a soft switch command from the scenario service layer or a hard switch command from the system service layer, and generating a motor control signal; receiving, by the atomic service layer, the motor control signal, and transmitting the motor control signal to a partition-glass control I/O abstraction service of the I/O abstraction layer for intervening in motor motion to control the opening and closing of a partition glass.
    Type: Application
    Filed: June 16, 2026
    Publication date: November 5, 2026
    Applicant: CHONGQING CHANGAN AUTOMOBILE CO., LTD.
    Inventors: Hao KANG, Xiaoqiang DONG
  • Publication number: 20260350374
    Abstract: A data processing system implements analyzing a textual input using an LLM to generate a query data matrix and a key data matrix; quantizing the query data matrix and the key data matrix using a quantization framework to generate a quantized query data matrix and a quantized key data matrix. Each channel in the query data matrix is associated with a channel-specific zero point value and each channel in the key data matrix is associated with a channel-specific zero point value. The system further implements providing the quantized matrices as an input to an attention mechanism of the LLM; analyzing the quantized matrices using the attention mechanism to determine attention scores without dequantizing data of the quantized matrices; and analyzing the attention scores using a softmax operation to generate a probability distribution that represents an attention weight associated with each key of the quantized key data matrix.
    Type: Application
    Filed: April 30, 2025
    Publication date: November 5, 2026
    Applicant: Microsoft Technology Licensing, LLC
    Inventors: Vedula Venkata Srikant BHARADWAJ, Saravanakumar RAJMOHAN, Victor Jonas RÜHLE, Hao KANG
  • Publication number: 20260332842
    Abstract: Aspects of the subject disclosure may include detecting uplink data packets on a network that conveys application protocol messages from an application on a mobile device to an application server providing a service to the application over the network; providing data packets to a machine learning model that is trained to identify respective application protocol messages; receiving from the machine learning model an indication that a start video download application protocol message has been identified and which requests initiation of a video download to the application from the application server; detecting downlink data packets on the network that conveys video content data to the mobile device in response to the start video download application protocol message; comparing a watch time for displaying the video content data on the mobile device by the application and a download data time; and determining a stall event has occurred based on the comparing.
    Type: Application
    Filed: July 1, 2026
    Publication date: October 22, 2026
    Applicant: AT&T Intellectual Property I, L.P.
    Inventors: Yuan Ding, Natalia Schenck, Vinay Sharma, Hao Kang
  • Publication number: 20260278874
    Abstract: A computing system including one or more processing devices configured to execute a face encoder including a text encoder, a face extraction model, and a token fusion model. At the face encoder, the one or more processing devices receive a text prompt and receive an input image depicting a face. At the text encoder, the one or more processing devices compute a text token sequence based at least on the text prompt. At the face extraction model, the one or more processing devices compute a face patch token sequence that encodes the face. At the token fusion model, the one or more processing devices compute a multimodal token sequence based at least on the text token sequence and the face patch token sequence. At a pretrained diffusion model, the one or more processing devices compute and output an output image conditioned on the multimodal token sequence.
    Type: Application
    Filed: March 25, 2025
    Publication date: September 17, 2026
    Inventors: Zichuan LIU, Qing YAN, Xin LU, Liming JIANG, Yumin JIA, Hao KANG, Efstathios FOTIADIS
  • Publication number: 20260278852
    Abstract: A computing system including one or more processing devices configured to receive a text prompt and receive an input image depicting a face. At a face encoder that includes a face vision transformer and a text-conditioned image encoder, the one or more processing devices compute a first face embedding at the face vision transformer based at least on the input image. The one or more processing devices compute a face token sequence at the text-conditioned image encoder based at least on a face identification text prompt and the first face embedding. At one or more text encoders, the one or more processing devices compute a text token sequence based at least on the text prompt. At a first diffusion model, the one or more processing devices compute and output an output image based at least in part on the face token sequence and the text token sequence.
    Type: Application
    Filed: March 25, 2025
    Publication date: September 17, 2026
    Inventors: Zichuan Liu, Qing Yan, Xin Lu, Liming Jiang, Yumin Jia, Hao Kang, Efstathios Fotiadis, Min Jin Chong
  • Publication number: 20260278752
    Abstract: A computing system is provided, including processing circuitry configured to receive an input image of an object and a text label corresponding to the input image. The processing circuitry is further configured to input the text label and input image to a vision-language model to generate a text description of features of the object and input the text description to a language model to generate a background prompt and target bounding box information. The processing circuitry is further configured to segment the input image to generate a target segmentation mask region in a shape of the object and generate a mask image including the target segmentation mask region and an infill region. The processing circuitry is further configured to inpaint the infill region based on the input image, the mask image, and the background prompt, to thereby generate the output image.
    Type: Application
    Filed: March 13, 2025
    Publication date: September 17, 2026
    Inventors: Yumin Jia, Hao Kang, Xin Lu, Qing Yan, Liming Jiang, Zichuan Liu
  • Patent number: 12720142
    Abstract: Aspects of the subject disclosure may include, for example, detecting a plurality uplink data packets on a network, the plurality of uplink data packets conveying application protocol messages from an application on a mobile device to an application server providing a service to the application on the mobile device over the network; providing data packets of the plurality uplink data packets to a machine learning model, the machine learning model trained to identify respective application protocol messages from the data packets; receiving from the machine learning model an indication that a start video download application protocol message has been identified by the machine learning model, the start video download application protocol message requesting initiation of a video download to the application on the mobile device from the application server; detecting a plurality of downlink data packets on the network, the plurality of downlink data packets conveying video content data to the mobile device in respo
    Type: Grant
    Filed: December 3, 2024
    Date of Patent: August 25, 2026
    Assignee: AT&T Intellectual Property I, L.P.
    Inventors: Yuan Ding, Natalia Schenck, Vinay Sharma, Hao Kang
  • Publication number: 20260220831
    Abstract: A computing system receives an input prompt and input images, encodes the input prompt into text embeddings, and extracts facial embeddings representing facial features in the one or more input images. The system further refines the facial embeddings to learn facial similarity features, and inputs a control image and the facial similarity features into a control network to generate features. Then the features, latent noise, and the text embeddings are inputted into a diffusion transformer to generate the synthesized image, and an output is generated based on the synthesized image.
    Type: Application
    Filed: April 25, 2025
    Publication date: July 30, 2026
    Inventors: Liming Jiang, Qing Yan, Xin Lu, Zichuan Liu, Yumin Jia, Hao Kang
  • Publication number: 20260203959
    Abstract: A computing system is provided for generating a synthesized image using subject-driven text-to-image generation. The computing system includes a processing circuitry and memory storing instructions that, when executed, cause the processing circuitry to receive one or more input images of a subject and a user prompt, generate a mosaic of two or more component images based on the one or more input images and the user prompt, generate a mask to mark a placeholder area in the mosaic to be inpainted, use a generative model to inpaint the placeholder area, extract the inpainted placeholder area to generate an extracted image of the subject, and output the extracted image of the subject as the synthesized image.
    Type: Application
    Filed: March 28, 2025
    Publication date: July 16, 2026
    Inventors: Hao KANG, Efstathios Fotiadis, Xin Lu, Yumin Jia, Liming Jiang, Qing Yan, Min Jin Chong, Zichuan Liu
  • Publication number: 20260172462
    Abstract: A processing system may obtain network performance data sets associated with a video communication session from a customer premises equipment, the network performance data sets including data sets for measurements of video uplink and downlink data rates. The processing system may apply at least one of the network performance data sets to at least one rule set implemented by the processing system to obtain a decision output indicative of whether an occurrence of: an uplink or downlink video freeze is detected. The processing system may detect the at least one of the uplink video freeze or the downlink video freeze in accordance with the decision output. The processing system may then perform at least one remedial action in response to the detecting of at least one of: the uplink video freeze or the downlink video freeze.
    Type: Application
    Filed: February 9, 2026
    Publication date: June 18, 2026
    Inventors: Hao Kang, Junping Zhang, Yuan DIng, Tzvi Chumash, Pranith Dachepally, Matthew Foust
  • Publication number: 20260157526
    Abstract: A double-state working kneeling chair has two forms: a kneeling mode and a seat mode. The double-state working kneeling chair includes a seat support and a kneeling support, where a sitting surface is arranged at an upper end of the seat support, and the sitting surface and the seat support are fixedly connected; a kneeling surface is fixedly connected to an upper surface of the kneeling support, one end of the kneeling support is hinged with a lower end of the seat support through a rotating mechanism, the kneeling support has a degree of freedom of rotating upward along the rotating mechanism, a clamping structure is arranged at the other end of the kneeling support, a clamping block is arranged at the upper end of the seat support, and the clamping structure matches the clamping block from bottom to top; and when the kneeling chair is in the kneeling mode, the kneeling support is placed in parallel to the ground, and at the moment, the kneeling surface is parallel to the ground.
    Type: Application
    Filed: December 25, 2024
    Publication date: June 11, 2026
    Inventor: Hao KANG
  • Publication number: 20260160115
    Abstract: Disclosed in the present application is a method for controlling a timing lock box. The method for controlling a timing lock box includes the following steps: receiving an expected unlocking moment set by a user, time information represented by the expected unlocking moment including year, month, day, hour, and minute information; locking a lock of the timing lock box; determining whether the expected unlocking moment is reached; and in response to reaching the expected unlocking moment, controlling the lock to be unlocked at the expected unlocking moment.
    Type: Application
    Filed: December 26, 2024
    Publication date: June 11, 2026
    Inventor: Hao KANG
  • Publication number: 20260162412
    Abstract: The present disclosure describes techniques for automatically identifying a checkpoint of a machine learning model for deployment. A plurality of checkpoints are generated during training the machine learning model. The machine learning model is trained on a set of training images. A plurality of subject images is generated by each of the plurality of checkpoints. Subject similarity and global difference between images in each pair of images are computed. Each pair of images comprises one of the set of training images and one of the plurality of subject images generated by each of the plurality of checkpoints. Image generation qualities of the plurality of checkpoints are evaluated based on the subject similarity and the global difference. The checkpoint of the machine learning model for deployment is automatically identified based on the evaluated image generation qualities.
    Type: Application
    Filed: December 6, 2024
    Publication date: June 11, 2026
    Inventors: Hao Kang, Xin Lu, Yumin Jia
  • Patent number: 12549616
    Abstract: A processing system may obtain network performance data sets associated with a video communication session from a customer premises equipment, the network performance data sets including data sets for measurements of video uplink and downlink data rates. The processing system may apply at least one of the network performance data sets to at least one rule set implemented by the processing system to obtain a decision output indicative of whether an occurrence of: an uplink or downlink video freeze is detected. The processing system may detect the at least one of the uplink video freeze or the downlink video freeze in accordance with the decision output. The processing system may then perform at least one remedial action in response to the detecting of at least one of: the uplink video freeze or the downlink video freeze.
    Type: Grant
    Filed: March 22, 2024
    Date of Patent: February 10, 2026
    Assignee: AT&T Intellectual Property I, L.P.
    Inventors: Hao Kang, Junping Zhang, Yuan Ding, Tzvi Chumash, Pranith Dachepally, Matthew Foust
  • Publication number: 20250301031
    Abstract: A processing system may obtain network performance data sets associated with a video communication session from a customer premises equipment, the network performance data sets including data sets for measurements of video uplink and downlink data rates. The processing system may apply at least one of the network performance data sets to at least one rule set implemented by the processing system to obtain a decision output indicative of whether an occurrence of: an uplink or downlink video freeze is detected. The processing system may detect the at least one of the uplink video freeze or the downlink video freeze in accordance with the decision output. The processing system may then perform at least one remedial action in response to the detecting of at least one of: the uplink video freeze or the downlink video freeze.
    Type: Application
    Filed: March 22, 2024
    Publication date: September 25, 2025
    Inventors: Hao Kang, Junping Zhang, Yuan DIng, Tzvi Chumash, Pranith Dachepally, Matthew Foust
  • Publication number: 20250269991
    Abstract: A teetering rotor hub is described. The teetering rotor hub may be able to couple to multiple rotor blades and to a rotor. The teetering rotor hub may be associated with an aerial system. The teetering rotor hub may be able to rotate about a rotor spin axis of an associated motor. The teetering rotor hub may couple to the rotor or shaft. In some embodiments, the rotor hub may couple to a pair of rotor blades. The teetering rotor hub may be able to teeter along a second axis that is offset from the rotor spin axis, where the offset may be between zero and ninety degrees. By varying a speed signal provided to the rotor, the rotor blades coupled to the rotor hub may be moved in various ways to generate flapping motions that are able to control the direction of movement of the aerial system.
    Type: Application
    Filed: February 26, 2024
    Publication date: August 28, 2025
    Inventors: John W. Gerdes, Hao Kang, John T. Hrynuk, Elias J. Rigas
  • Publication number: 20250097503
    Abstract: Aspects of the subject disclosure may include, for example, detecting a plurality uplink data packets on a network, the plurality of uplink data packets conveying application protocol messages from an application on a mobile device to an application server providing a service to the application on the mobile device over the network; providing data packets of the plurality uplink data packets to a machine learning model, the machine learning model trained to identify respective application protocol messages from the data packets; receiving from the machine learning model an indication that a start video download application protocol message has been identified by the machine learning model, the start video download application protocol message requesting initiation of a video download to the application on the mobile device from the application server; detecting a plurality of downlink data packets on the network, the plurality of downlink data packets conveying video content data to the mobile device in respo
    Type: Application
    Filed: December 3, 2024
    Publication date: March 20, 2025
    Applicant: AT&T Intellectual Property I, L.P.
    Inventors: Yuan Ding, Natalia Schenck, Vinay Sharma, Hao Kang
  • Patent number: 12200286
    Abstract: Aspects of the subject disclosure may include, for example, detecting a plurality uplink data packets on a network, the plurality of uplink data packets conveying application protocol messages from an application on a mobile device to an application server providing a service to the application on the mobile device over the network; providing data packets of the plurality uplink data packets to a machine learning model, the machine learning model trained to identify respective application protocol messages from the data packets; receiving from the machine learning model an indication that a start video download application protocol message has been identified by the machine learning model, the start video download application protocol message requesting initiation of a video download to the application on the mobile device from the application server; detecting a plurality of downlink data packets on the network, the plurality of downlink data packets conveying video content data to the mobile device in respo
    Type: Grant
    Filed: September 2, 2022
    Date of Patent: January 14, 2025
    Assignee: AT&T Intellectual Property I, L.P.
    Inventors: Yuan Ding, Natalia Schenck, Vinay Sharma, Hao Kang
  • Patent number: 11932428
    Abstract: A projectile-launched aircraft system includes a projectile launcher including a triggering mechanism, a rotary-wing, hover-capable aircraft including a rotor assembly that includes at least one rotor blade, wherein the rotor blade includes a stowed configuration and a deployed configuration that is circumferentially spaced from the stowed configuration about a pivot axis, wherein, upon actuation of the triggering mechanism, the projectile launcher is configured to launch the aircraft along a flightpath.
    Type: Grant
    Filed: May 22, 2020
    Date of Patent: March 19, 2024
    Assignee: The Texas A&M University System
    Inventors: Moble Benedict, Hunter J. Denton, Hao Kang, Vikram Hrishikeshavan