Abstract: The method includes fine-tuning a pre-trained audio encoder on supervised speech recognition training data. For each transcribed speech utterance, the method includes processing a corresponding sequence of audio features to generate a sequence of audio encoder posteriors over a first vocabulary of output labels using the fine-tuned audio encoder, determining a sequence of speech embeddings by computing a weighted sum of an input embedding table of a pre-trained LLM from the sequence of audio encoder posteriors, processing a concatenation of the sequence of speech embeddings and a sequence of text embeddings representative of a corresponding ground-truth transcription to generate a predicted sequence of output labels by the pre-trained LLM, and determining a cross-entropy loss term based on the predicted sequence of output labels and the ground-truth transcription. The method includes fine-tuning the pre-trained LLM based on each cross-entropy loss term.

Abstract: A method of a multilingual ASR model includes receiving a sequence of acoustic frames characterizing an utterance of speech. At a plurality of output steps, the method further includes generating a first higher order feature representation for an acoustic frame by a first encoder that includes a first plurality of multi-head attention layers; generating a second higher order feature representation for a corresponding first higher order feature representation by a second encoder that includes a second plurality of multi-head attention layers; and generating, by a first decoder, a first probability distribution over possible speech recognition hypotheses based on the second higher order feature representation and a sequence of N previous non-blank symbols. A gating layer of each respective MoE layer configured to dynamically route an output from a previous multi-head attention layer at each of the plurality of output steps to a respective pair of feed-forward expert networks.

Abstract: The present disclosure is generally directed to a modular busbar assembly that can be adjusted to fit varying rack power targets and connector interfaces, accommodating multiple connector geometries along one cost-effective assembly. The modular busbar assembly enables high power densities at powershelf regions by accommodating single-and double-pole connectors and connector contacts at various positions relative to the modular busbar assembly. The modular busbar assembly is customizable because the auxiliary busbar geometry may be altered independently from the primary busbar geometry.

Abstract: A method includes obtaining a plurality of contextual data elements associated with a user of a user device. For each respective contextual data element, the method includes determining a corresponding weight indicating an amount of relevance of the respective contextual data element to the user. The method includes determining a context based on the corresponding weight of each respective contextual data element. The context includes a portion of the contextual data elements from the plurality of contextual data elements. The method includes sending the context to the user device. The user device is configured to store the context locally at the user device based on receiving the context.

Abstract: Techniques and apparatuses are described that perform gesture-based control using active acoustic sensing. By transmitting and receiving acoustic signals, a hearable can recognize changes in an acoustic circuit to perform gesture recognition. Gesture recognition involves recognizing a muscle-based gesture and/or an object-based gesture. With gesture recognition, the hearable can support a voice-free and/or hands-free user interface that enables the user to control an operation of the hearable or a computing device that is coupled to the hearable. This voice-free and/or hands-free user interface can provide a discreet and socially acceptable means of controlling an electronic device in a variety of different environments. It can also provide additional accessibility for people with various disabilities or physical restrictions.

Abstract: A method includes receiving a query and context associated with the query. The method includes determining one or more context components from the context. For each respective context component of the one or more context components, the method includes determining a corresponding priority score based on a relevance of the respective context component to the query and biasing the respective context component based on the corresponding priority score. The method includes generating, using a neural network model, a response based on the query and the biased one or more context components.

Abstract: A method includes receiving a first query issued by a user and processing the first query to classify the first query as being related to a particular existing topic that corresponds to a respective one of a plurality of topic summaries stored in a topic summary datastore. Each respective topic summary of the plurality of topic summaries stored in the topic summary datastore corresponds to a different respective topic and is associated with a respective summary of past query-response interactions between a user and an assistant interface that are related to the different respective topic. The method also includes retrieving the respective topic summary from the topic summary datastore that corresponds to the particular existing topic, processing the first query conditioned on the respective topic summary retrieved from the topic summary datastore to generate a first response, and providing presentation content based on the first response for output.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using frames for performing tasks. One of the methods includes receiving a first request to perform a task, the first request comprising user speech identifying the task; generating a frame associated with the task, wherein the frame comprises one or more types of values necessary to perform the task, and wherein each type of value can be satisfied by a respective value; receiving a second request to provide information related to a question, the second request comprising user speech identifying the question; providing information identifying the question to a search engine, and receiving a response identifying one or more terms; determining that at least one term can satisfy a type of value necessary to perform the task; and storing the at least one term in the frame.

Abstract: Implementations for a functional unit are provided, wherein the functional unit can accumulate more than two serial inputs and provide one serial summation output. The serial inputs and outputs can be single bits or multiple bit busses. The functional unit can be implemented as a logical tree, where any two points are connected by one path. The functional unit can be incorporated into a processing unit of a programmable device to allow for construction of various functions.

Abstract: A method for managing collaborative playlists is disclosed. The method includes providing, by a processing device, indication regarding one or more available collaborative guidance controls, receiving, by the processing device, first user selection regarding one or more of the collaborative guidance controls, receiving, at the processing device, second user suggestion regarding an additional content item for addition to the playlist, and responsive to determining that the second user suggestion satisfies the one or more collaborative guidance controls, adding the additional content item to the playlist for consumption by the first user and a plurality of second users via a plurality of user devices.

Abstract: The present disclosure describes various aspects of radio link management to enable unpaired receiver paths of user equipment. In some aspects, a modem of a user equipment is coupled with multiple receiver paths that include an unpaired receiver path. To enable use of the unpaired receiver path in various modes of multi-channel communication, a radio link manager of the user equipment modifies configuration information of the modem to add a nonexistent receiver path to a set of receiver path parameters for a frequency band in which the unpaired receiver path is capable of operating. Along with adding the nonexistent receiver path, the radio link manager may also modify the receiver path parameters to enable the unpaired receiver path for the frequency band. The nonexistent receiver path and unpaired receiver path are then exposed for use by the modem to implement one or more multi-channel receive modes for the frequency band.

Abstract: Implementations relate to handling visual content across a multi-turn dialog. A user input that includes natural language content and visual content is received during the dialog. If the visual content is being received for the first time in the dialog, the visual content is processed to generate a corresponding tokenized representation of the visual content. The corresponding tokenized representation can be cached in a database in association with the dialog, or in association with a user account of a user of the user query. If the visual content is subsequently referenced in the dialog, the corresponding tokenized representation of the visual content is retrieved from the database. The corresponding tokenized representation of the visual content, corresponding tokenized representations of natural language content, and optionally other metadata can be processed, using a generative model, to generate a response responsive to the user input.

Abstract: A method for no-code data driven workflows using external data triggers includes generating a no-code application that includes a trigger condition associated with a first data source and an action response defining an action in response to the trigger condition being satisfied. The method includes receiving a deployment request requesting deployment of the no-code application within an application environment. After receiving the deployment request, the method includes receiving an indication indicating that the trigger condition is satisfied based on trigger data stored at the first data source. The indication is associated with a user of the no-code application. Based on receiving the indication, the method includes executing the action response. After executing the action response, the method includes receiving action response data. The method also includes storing the action response data at a second data source and notifying the user using the action response data.

Abstract: Implementations relate to storing historical queries processed using a generative model in association with intermediate model outputs generated using the generative model for each of the historical queries. Implementations further relate to receiving a user query processable using the generative model. In response to receiving the user query, the user query can be compared to the historical queries to identify a particular historical query (e.g., having a similarity score satisfying a similarity threshold) that matches the user query. Particular intermediate model output associated with the particular historical query can be selected from all intermediate model outputs stored in association with the particular historical query, and a response to the user query can be generated based at least on the selected particular intermediate model output associated with the particular historical query.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for providing knowledge panels with search results. In one aspect, a method includes receiving a query. Search results that are responsive to the query are obtained. A first set of factual entities referenced by the query are identified. A particular factual entity is selected from the first set. It is determined that a knowledge panel for the particular factual entity is to be provided with the search results based, at least in part, on content of the knowledge panel and characteristics of the search results. The knowledge panel can include at least one content item, for the selected factual entity, received from a first resource and at least one content item, for the selected factual entity, received from a second resource different than the first resource. The knowledge panel is provided.

Abstract: Systems and methods for generating a domain-specific conversational automated assistant. In some examples, a conversational language model is used to generate a target answer and a target action recommendation in response to each of a set of in-domain training questions. In some examples, the conversational language model is further used to generate follow-up questions to one or more of its generated target answers, and to then generate a target answer and target action recommendation to each generated follow-up question. In some examples, the processing system also generates a set of out-of-domain training examples including an out-of-domain question, a predetermined target answer, and a predetermined target action recommendation. The automated assistant may then be trained to predict the generated target answers and target action recommendations based on the associated training question or generated follow-up question, as well as any prior questions and answers in the conversation.

Abstract: This document describes a biometric authentication system including a biometric authentication manager is configured to receive user input at a touch-sensitive display, determine a shape of the user input, and alter a luminosity and/or color of one or more portions of a region of the touch-sensitive display. Through such a technique, the biometric authentication manager can implement spatially, temporally, and/or chromatically variable illumination patterns to produce a fingerprint reflection, receive the fingerprint reflection, and then analyze the fingerprint reflection for authenticity and/or for liveness characteristics effective to detection spoof fingerprints.

Abstract: Capture information is caused to be sent to a user computing device, wherein the capture information comprises instructions to a user of the user computing device to capture a plurality of image sets for a Point of Interest (POI). The plurality of image sets for the POI are received, wherein the plurality of image sets comprises a calibration image set, a POI depiction image set, and a path image set. A machine-learned POI representation model is trained with the plurality of image sets for the POI, wherein the machine-learned POI representation model is trained to generate a representation of the POI. The machine-learned POI representation model is used to generate video data that depicts a virtual camera traversing a path from a starting location to an ending location within the POI.

Abstract: A UE receives satellite ephemeris information from a server and calculates one or more satellite parameters for at least one satellite based on the satellite ephemeris information. The UE controls, based on the one or more satellite parameters, operation of a software application that utilizes a data service provided via the at least one satellite. Calculating satellite parameters can include calculating one or more satellite parameters at an API of the UE, and controlling operation of a software application can include receiving, at the API, a request for satellite ephemeris information from the software application, providing, by the API, a representation of the one or more satellite parameters to the software application in response to the request, and adjusting, at the software application, an operation of the software application based on the received representation of the one or more satellite parameters.

Abstract: A circuit device includes a plurality of driver circuits, each driver circuit configured to output a respective current, and a switch network including a plurality of switch network inputs. Each switch network input of the plurality of switch network inputs is coupled to an output of a respective driver circuit of the plurality of driver circuits to receive the respective current from the respective driver circuit, a switch network output, one or more control inputs, and a plurality of switches. The switch network is configured to adjust one or more switches of the plurality of switches to provide, to a quantum computing circuit at the switch network output, an output signal including a sum of a subset of the currents from the plurality of driver circuits, the subset of the currents corresponding to the one or more control signals.