Abstract: The present application provides a method for preparing a perovskite thin film and a perovskite cell. The perovskite cell includes a cell baseplate, a first electrode layer, a first carrier transport layer, a carrier transport passivation layer, a perovskite thin film, a second carrier transport layer, and a second electrode layer. In the present application, by adding a thickener into a perovskite inorganic salt solution, fluid characteristics of the perovskite inorganic salt solution can be effectively improved, resulting in a high-viscosity first inorganic substrate layer. The fluidity of the first inorganic substrate layer decreases as the temperature decreases, thereby enabling stable deposition and adhesion of the first inorganic substrate layer on a cell substrate.

Abstract: Disclosed are systems and techniques that may generate prompts for language models. The techniques include obtaining a first dataset and a second dataset and training a hierarchical virtual token generator (VTG) model to generate a large language model (LLM) input prompt. Training the hierarchical VTG includes training, based on the first dataset, a first VTG to output a first virtual token and training, based on the second dataset, a second VTG to output a second virtual token embedding. The generated LLM input prompt includes the first virtual token embedding and the second virtual token embedding.

Abstract: Systems and methods are described for detecting changes at a location based on image data by a mobile robot. A system can instruct navigation of the mobile robot to a location. For example, the system can instruct navigation to the location as part of an inspection mission. The system can obtain input identifying a change detection. Based on the change detection and obtained image data associated with the location, the system can perform the change detection and detect a change associated with the location. For example, the system can perform the change detection based on one or more regions of interest of the obtained image data. Based on the detected change and a reference model, the system can determine presence of an anomaly condition in the obtained image data.

Abstract: Provided are a high-strength and high-plasticity hot-forming steel with oxidation resistance for automobiles and a hot-forming process thereof, and the hot-forming steel has chemical compositions in mass percentages as follows: C: 0.35%-50%, Si: ?0.20%, Mn: 1.50%-2.50%, P: 0.050%-0.10%, S?0.004%, Als: 0.02%-0.06%, Nb: 0.03%-0.07%, Ti: 0.020%-0.050%, V: 0.08%-0.15%, Cr: 1.50%-3.20%, Mo: 0.10%-0.30%, B: ?0.0040%, N?0.005%, the balance Fe and inevitable impurities. The hot-forming steel provided in the present invention has high oxidation resistance, high strength and plasticity, does not need atmosphere protection during hot forming, and does not need shot blasting treatment after hot forming.

Abstract: The present disclosure discloses a rare earth permanent magnet, its preparation method and motors. The rare earth permanent magnet comprises a light rare earth element and a heavy rare earth element. The light rare earth element must contain Nd and the heavy rare earth element must contain Dy and Tb; along the width direction, the rare earth permanent magnet has two edge portions and one middle portion; along the direction from the outer edge of the edge portion towards the central axis of the middle portion, the weight percentage of Dy gradually increases, while the weight percentage of Tb gradually decreases; the average coercive force of the edge portion is more than that of the middle portion. The rare earth permanent magnet of the present disclosure has a good anti-demagnetization performance.

Abstract: Methods and apparatus for operating a mobile robot in a loading dock environment are provided. The method comprises capturing, by a camera system of the mobile robot, at least one image of the loading dock environment, and processing, by at least one hardware processor of the mobile robot, the at least one image using a machine learning model trained to identify one or more features of the loading dock environment.

Abstract: Disclosed are systems and techniques for training machine learning models. The techniques include providing a first data of a first modality as input to a first machine learning model to obtain a first output of a second modality, providing the first output of the second modality as input to a second machine learning model to obtain a second output of the first modality, providing the first data as input to a third machine learning model to obtain a first tensor, providing the second output as input to the third machine learning model to obtain a second tensor, calculating a first loss based on a comparison between the first tensor and the second tensor, and causing the first machine learning model to be modified based on the first loss.

Abstract: Apparatuses, systems, and techniques are presented to generate digital models. In at least one embodiment, a neural network is used to generate motions of a first portion of an object based on motions of a second portion of the object and audio corresponding to the motions of the second portion of the object. For example, a neural network may use audio information to generate motions of facial and body features of a digital model.

Abstract: Apparatuses, systems, and techniques to generate software code. In at least one embodiment, one or more neural networks are to use information indicating one or more software program dependencies to generate one or more software programs having said indicated one or more software program dependencies.

Abstract: In various examples, systems and methods are disclosed relating to engaging users through a personalized interface. One system includes at least one processor configured to determine first conversation history with a user. The at least one processor further configured to determine a response by applying the first conversation history with the user to a machine learning model, wherein the machine learning model is updated using user input indicative of an interest level of the user for each of a plurality of candidate responses to a question, a content of the question is determined by the at least one processor, and the plurality of candidate responses are determined using the machine learning model, and the machine learning model is updated using the user input as a reward signal.

Abstract: The present invention provides a WAPI trusted wireless local area network device, and relates to the field of WAPI local area network devices. The WAPI trusted wireless local area network device includes a data obtaining module, an information obtaining module, a key module, an encryption module, a data transmission module, a data receiving module, a decoding module, a storage module, a data check module, an alarm module, and a data output module. Different keys are generated based on a sending time of initial data, primary encryption is performed on the initial data by using a key, and then secondary encryption is performed based on a preset encryption protocol. The key has a self-updating ability according to the time, which can make an encryption manner of each group of initial data different, and improve the security of data transmission.

Abstract: Disclosed herein is a system that comprises an imaging sensor coupled to a robot. The system also comprises an imaging module configured to capture images of a scene within a field of view of the imaging sensor, which is coupled to the robot. The system further comprises an image-processing module configured to process the captured images to identify one or more objects within the field of view of the imaging sensor in the scene relative to the robot. The system additionally comprises a tracking module configured to track the one or more objects relative to the robot while the one or more objects are in motion based on the captured images.

Abstract: The present invention provides a WAPI trusted wireless local area network device, and relates to the field of WAPI local area network devices. The WAPI trusted wireless local area network device includes a data obtaining module, an information obtaining module, a key module, an encryption module, a data transmission module, a data receiving module, a decoding module, a storage module, a data check module, an alarm module, and a data output module. Different keys are generated based on a sending time of initial data, primary encryption is performed on the initial data by using a key, and then secondary encryption is performed based on a preset encryption protocol. The key has a self-updating ability according to the time, which can make an encryption manner of each group of initial data different, and improve the security of data transmission.

Abstract: Disclosed are systems and techniques for training personalized language models. The techniques include applying a plurality of first machine learning models to a first input prompt. Each of the plurality of first machine learning models generates a respective reward value of a first plurality of reward values. The techniques include applying a second machine learning model to the first plurality of reward values to obtain first reward value embeddings; applying a third machine learning model to the first reward value embeddings and the first input prompt to obtain a first output response; calculating a first loss based on a comparison between the first output response and the first input prompt; and causing the second machine learning model to be modified based on the first loss.

Abstract: Disclosed are systems and techniques for training machine learning models. The techniques include providing a first data of a first modality as input to a first machine learning model to obtain a first output of a second modality, providing the first output of the second modality as input to a second machine learning model to obtain a second output of the first modality, providing the first data as input to a third machine learning model to obtain a first tensor, providing the second output as input to the third machine learning model to obtain a second tensor, calculating a first loss based on a comparison between the first tensor and the second tensor, and causing the first machine learning model to be modified based on the first loss.

Abstract: In various examples, techniques for accelerating inference in text and speech processing for conversational AI systems and applications is described herein. Systems and methods are disclosed that use one or more techniques, such as token merging, in order to reduce a number of tokens processed by one or more machine learning models. For instance, the machine learning model(s) may process text and, based at least on the processing, generate scores (e.g., attention scores) indicating relationships between tokens associated with the text. The machine learning model(s) may then use the scores to merge at least one pair of the tokens. As described herein, the merging may reduce the overall number of tokens associated with the text while still maintaining the same semantic meaning as the original text. Next, the machine learning model(s) may process the reduce number of tokens in order to determine an output associated with the text.

Abstract: Systems and methods are described for detecting changes at a location based on image data by a mobile robot. A system can instruct navigation of the mobile robot to a location. For example, the system can instruct navigation to the location as part of an inspection mission. The system can obtain input identifying a change detection. Based on the change detection and obtained image data associated with the location, the system can perform the change detection and detect a change associated with the location. For example, the system can perform the change detection based on one or more regions of interest of the obtained image data. Based on the detected change and a reference model, the system can determine presence of an anomaly condition in the obtained image data.

Abstract: Disclosed herein is a system that comprises an imaging sensor coupled to a robot. The system also comprises an imaging module configured to capture images of a scene within a field of view of the imaging sensor, which is coupled to the robot. The system further comprises an image-processing module configured to process the captured images to identify an object above a horizon in the scene within the field of view of the imaging sensor. The system additionally comprises a tracking module configured to track a trajectory of the object based on the captured images to determine positions of the object relative to the robot.

Abstract: Disclosed are systems and techniques that may generate prompts for language models. The techniques include obtaining a first dataset and a second dataset and training a hierarchical virtual token generator (VTG) model to generate a large language model (LLM) input prompt. Training the hierarchical VTG includes training, based on the first dataset, a first VTG to output a first virtual token and training, based on the second dataset, a second VTG to output a second virtual token embedding. The generated LLM input prompt includes the first virtual token embedding and the second virtual token embedding.

Abstract: Methods and apparatus for operating a mobile robot in a loading dock environment are provided. The method comprises capturing, by a camera system of the mobile robot, at least one image of the loading dock environment, and processing, by at least one hardware processor of the mobile robot, the at least one image using a machine learning model trained to identify one or more features of the loading dock environment.