Abstract: Implementations relate to generating standardized metrics from device specific metrics that are generated during an interaction between a user and an automated assistant. The metrics indicate events that occurred while processing an interaction of a user with the automated assistant and are specific to the particular configuration of the device with which the user is interacting. Conversion mappings are determined based on device characteristics that can be utilized to convert the device metrics into standardized metrics. Analysis metrics are generated based on the standardized metrics that are incapable of being generated from the device metrics. Some implementations include visually rendering the analysis metrics such that one or more of the analysis metrics are rendered more prominently than other metrics.

Abstract: The present disclosure is generally related to a data processing system to detect potential exfiltration of audio data by agent applications can include a data processing system. The data processing system can identify, from an I/O record, an input received from the digital assistant application via a microphone of a client device, an output received from the agent application after the input, and a microphone status for the microphone. The data processing system can determine that the output is terminal based on the input and the output. The data processing system can identify the microphone status as in the enabled state subsequent to the input. The data processing system can determine that the agent application is unauthorized to access audio data acquired via the microphone of the client device based on determining that the output is terminal and identifying the microphone status as enabled.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for providing suggestions within a document. In one aspect, a method includes obtaining textual input provided to a document editing application by a user device, the textual input being provided to the document editing application for inclusion in a document; identifying performance measures associated with the current editing session for the document, each performance measure being based on session data obtained from the user device during a document editing session, the session data being for the textual input and prior text that was included in the document prior to the textual input; providing the performance measures as input to a suggestion model that was trained using historical performance measures identified in performance logs for historical document editing sessions of users; and throttling textual suggestions during the current editing session based on the output of the suggestion model.

Abstract: Systems and methods for manipulation of the polarization state of light emitted by a laser projector to reduce the angle range of a scanning mirror articulated by a micro-electromechanical system MEMS to reduce power consumption are disclosed. A system includes a light source configured to emit laser light, a scanning mirror, and an angle expander disposed between the light source and the scanning mirror, the angle expander being configured to cause the laser light from the light source to be reflected at least once from the angle expander and at least twice from the scanning mirror.

Abstract: Methods, systems, and media for ambient background noise modification are provided. In some implementations, the method comprises: identifying at least one noise present in an environment of a user having a user device, an activity the user is currently engaged in, and a physical or emotional state of the user; determining a target ambient noise to be produced in the environment based at least in part on the identified noise, the activity the user is currently engaged in, and the physical or emotional state of the user; identifying at least one device associated with the user device to be used to produce the target ambient noise; determining sound outputs corresponding to each of the one or more identified devices, wherein a combination of the sound outputs produces an approximation of one or more characteristics of the target ambient noise; and causing the one or more identified devices to produce the determined sound outputs.

Abstract: Example aspects of the present disclosure are directed to systems and methods that learn a compressed representation of a machine-learned model (e.g., neural network) via representation of the model parameters within a reparameterization space during training of the model. In particular, the present disclosure describes an end-to-end model weight compression approach that employs a latent-variable data compression method. The model parameters (e.g., weights and biases) are represented in a “latent” or “reparameterization” space, amounting to a reparameterization. In some implementations, this space can be equipped with a learned probability model, which is used first to impose an entropy penalty on the parameter representation during training, and second to compress the representation using arithmetic coding after training. The proposed approach can thus maximize accuracy and model compressibility jointly, in an end-to-end fashion, with the rate-error trade-off specified by a hyperparameter.

Abstract: In one example, a method includes method comprising: receiving audio data generated by a microphone of a current computing device; identifying, based on the audio data, one or more computing devices that each emitted a respective audio signal in response to speech reception being activated at the current computing device; and selecting either the current computing device or a particular computing device from the identified one or more computing devices to satisfy a spoken utterance determined based on the audio data.

Abstract: A computer-implemented method to perform image-to-image translation. The method can include obtaining one or more machine-learned generator models. The one or more machine-learned generator models can be configured to receive an input image and a user-specified conditioning vector that parameterizes one or more desired values for one or more defined characteristics of an output image. The one or more machine-learned generator models can be configured to perform, based at least in part on the user-specified conditioning vector, one or more transformations on the input image to generate the output image with the one or more desired values for the one or more defined characteristics. The method can include receiving the input image and the user-specified conditioning vector. The method can include generating, using the machine-learned generator model, an output image having the one or more desired values for the one or more characteristics.

Abstract: Systems and methods for deploying countermeasures against unauthorized scripts interfering with the rendering of content elements on information resources are provided herein. A computing device can receive an information resource including a content rendering verification script and a first content element. The computing device can execute the script. The computing device can render the first content element for display on the information resource in a first format. The computing device can determine that the first content element is not successfully displayed in the first format. The computing device can render the first content element for display on the information resource in a second format, responsive to the determination. The computing device can determine that the first content element is successfully displayed in the second format. The computing device can display a second content element of the information resource responsive to the determination.

Abstract: Implementations relate to generating multi-modal response(s) through utilization of large language model(s) (LLM(s)). Processor(s) of a system can: receive natural language (NL) based input, generate a multi-modal response that is responsive to the NL based output, and cause the multi-modal response to be rendered. In some implementations, and in generating the multi-modal response, the processor(s) can process, using a LLM, LLM input (e.g., that includes at least the NL based input) to generate LLM output, and determine, based on the LLM output, textual content for inclusion in the multi-modal response and multimedia content for inclusion in the multi-modal response. In some implementations, the multimedia content can be obtained based on a multimedia content tag that is included in the LLM output and that is indicative of the multimedia content. In various implementations, the multimedia content can be interleaved between segments of the textual content.

Abstract: Implementations set forth herein relate to conditionally caching responses to automated assistant queries according to certain contextual data that may be associated with each automated assistant query. Each query can be identified based on historical interactions between a user and an automated assistant, and—depending on the query, fulfillment data can be cached according to certain contextual data that influences the query response. Depending on how the contextual data changes, a cached response stored at a client device can be discarded and/or replaced with an updated cached response. For example, a query that users commonly ask prior to leaving for work can have a corresponding assistant response that depends on features of an environment of the users. This unique assistant response can be cached, before the users provide the query, to minimize latency that can occur when network or processing bandwidth is unpredictable.

Abstract: The present disclosure is generally directed to reconstructing representations of bodies from images. An example method of the present disclosure includes inputting, into a machine-learned reconstruction model, input data descriptive of an image depicting a body; predicting, using a machine-learned marker prediction component of the reconstruction model, a set of surface marker locations on the body; and outputting, using a machine-learned marker poser component of the reconstruction model, an output representation of the body that corresponds to the set of surface marker locations. In the example method, one or more parameters of the reconstruction model were learned at least in part based on a consistency loss corresponding to a distance between relaxed-constraint representations generated from a prior set of surface marker locations predicted according to the one or more parameters and parametric representations generated from the prior set using kinematic constraints associated with the body.

Abstract: Some implementations are directed to generating a personal database entry for a user based on free-form natural language input formulated by the user via one or more user interface input devices of a computing device of the user. The generated personal database entry may include one or more terms of the natural language input and descriptive metadata determined based on one or more terms of the natural language input and/or based on contextual features associated with receiving the natural language input. Some implementations are directed to generating, based on one or more personal database entries of a user, output that is responsive to further free-form natural language input of the user. For example, one or more entries that are responsive to further natural language input of the user can be identified based on matching content of those entries to one or more search parameters determined based on the further input.

Abstract: Techniques are disclosed that enable model predictive control of a robot based on a latent dynamics model and a reward function. In many implementations, the latent space can be divided into a deterministic portion and stochastic portion, allowing the model to be utilized in generating more likely robot trajectories. Additional or alternative implementations include many reward functions, where each reward function corresponds to a different robot task.

Abstract: To control transmissions to a user equipment (UE) over a downlink (DL) multiple-input, multiple-output (MIMO) channel, a base station determining that the UE is configured to support N DL MIMO layers and transmit reference signals over L antenna chains (852). In response to determining that L<N (870), the base station generates channel information for the DL MIMO channel using (i) L uplink reference signals, each transmitted by the UE over a respective one of the L antenna chains (880), and (ii) one or more additional transmissions received from the UE (882). The base station transmits data streams over the DL MIMO channel in accordance with the generated channel information.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for handing off a user conversation between computer-implemented agents. One of the methods includes receiving, by a computer-implemented agent specific to a user device, a digital representation of speech encoding an utterance, determining, by the computer-implemented agent, that the utterance specifies a requirement to establish a communication with another computer-implemented agent, and establishing, by the computer-implemented agent, a communication between the other computer-implemented agent and the user device.

Abstract: A two-pass encoding operation is implemented to encode one or more gaming frames into a game stream. The two-pass encoding operation includes a first encoding pass performed on a current frame. As a result of the first encoding pass, an estimated complexity for the current frame is determined. The resulting estimated complexity is then modulated according to a quality difference between reference frames used during the first pass encoding and a subsequent second pass encoding. Based on the modulated complexity, a quantization parameter is determined for the current frame that is then used to perform a second pass encoding on the current frame, resulting in an encoded frame. This encoded frame is then transmitted as part of a stream to a client system.

Abstract: Implementations are directed to updating a trained voice bot that is deployed for conducting conversations on behalf of a third-party. A third-party developer can interact with a voice bot development system that enables the third-party developer to train, update, validate, and monitor performance of the trained voice bot. In various implementations, the trained voice bot can be updated by updating a corpus of training instances that was initially utilized to train the voice bot, and updating the trained voice bot based on the updated corpus. In some implementations, the corpus of training instances may be updated in response to identifying occurrence(s) of behavioral error(s) of the trained voice bot while the conversations are being conducted on behalf of the third-party. In additional or alternative implementations, the corpus of training instances may be updated in response to determining the trained voice bot does not include a desired behavior.

Abstract: Implementations utilize deep reinforcement learning to train a policy neural network that parameterizes a policy for determining a robotic action based on a current state. Some of those implementations collect experience data from multiple robots that operate simultaneously. Each robot generates instances of experience data during iterative performance of episodes that are each explorations of performing a task, and that are each guided based on the policy network and the current policy parameters for the policy network during the episode. The collected experience data is generated during the episodes and is used to train the policy network by iteratively updating policy parameters of the policy network based on a batch of collected experience data. Further, prior to performance of each of a plurality of episodes performed by the robots, the current updated policy parameters can be provided (or retrieved) for utilization in performance of the episode.

Abstract: At least selectively utilizing a large language model (LLM) in generating a natural language (NL) based summary to be rendered in response to a query. In some implementations, in generating the NL based summary additional content is processed using the LLM. The additional content is in addition to query content of the query itself and, in generating the NL based summary, can be processed using the LLM and along with the query content—or even independent of the query content. Processing the additional content can, for example, mitigate occurrences of the NL based summary including inaccuracies and/or can mitigate occurrences of the NL based summary being over-specified and/or under-specified.