Abstract: Certain aspects of the disclosure provide techniques for delivering consistent data in an eventually consistent system. An example method includes storing a first data table comprising a trigger timestamp; storing a second data table comprising a commit timestamp and an adaptor offset corresponding to a latest capture change data (CDC) event recorded in a categorized data stream; determining that the commit timestamp is later than or the same as the trigger timestamp; storing a third data table comprising a file writer offset corresponding to a latest data file stored in a file storage system; determining that the file writer offset is greater than or equal to the adaptor offset; causing the data files stored in the file storage system to be transmitted to and stored in a data lake; and generating and transmitting a notification comprising an indication that the data lake is up-to-date as of the trigger timestamp.

Abstract: Systems and methods are disclosed for a resource efficient partial bootstrap of one or more databases. Instead of the typical requirement of deleting all data from a destination database and performing a full bootstrap when deletions in data at a data source occurs, implementations herein disclose a resource efficient partial bootstrap process to identify deletions in data at the data source and update the destination database accordingly without requiring a full bootstrap. To perform partial bootstrapping, a system compares the set of data keys from a source index at the data source with the set of data keys from a destination index at the destination database to identify a difference between the two sets of data keys. The system then deletes the data keys in the destination index that do not appear in the source index (as identified in the difference between the two sets of data keys).

Abstract: This disclosure describes a framework for removing or mitigating inconsistency bias in generative artificial intelligence (AI) model responses, which inherently provide generative outputs that may include biases for certain groups. Specifically, this disclosure describes a model bias removal system (e.g., a model inconsistency bias mitigation system) that influences a generative AI model to respond to user prompts without inconsistency biases while not influencing or affecting other aspects of the model's ability to generate user responses. By doing so, the model bias removal system improves the accuracy and efficiency of generative AI models. Additionally, the model bias removal system enhances the fairness, consistency, and impartiality of generative AI model responses.

Abstract: The technology described herein, among other things, relates to testing applications, backed by language models (LMs), for compliance with responsible artificial-intelligence (RAI) guidelines. For example, LM-based chatbots have proliferated across many different domains and implementations. These chatbots, however, may be susceptible to attacks or attempts to cause the chatbots to violate RAI guidelines by producing harmful content and/or potentially violating copyrights. To evaluate whether an LM-based application, such as a chatbot, is complying with respective RAI guidelines, the technology disclosed herein adaptively simulates conversations with the LM-based application in an attempt to cause the LM-based application to violate the RAI guidelines in a controlled, simulated environment.

Abstract: An example computing device may include memory and one or more processors. The one or more processors may be configured to parallel entropy decode encoded video data from a received bitstream to generate entropy decoded data. The one or more processors may be configured to predict a motion vector based on the entropy decoded data. The one or more processors may be configured to decode a motion vector residual from the entropy decoded data. The one or more processors may be configured to add the motion vector residual and motion vector. The one or more processors may be configured to warp previous reconstructed video data with an overlapped block-based warp function using the motion vector to generate predicted current video data. The one or more processors may be configured to sum the predicted current video data with a residual block to generate current reconstructed video data.

Abstract: Systems and methods are disclosed for a resource efficient partial bootstrap of one or more databases. Instead of the typical requirement of deleting all data from a destination database and performing a full bootstrap when deletions in data at a data source occurs, implementations herein disclose a resource efficient partial bootstrap process to identify deletions in data at the data source and update the destination database accordingly without requiring a full bootstrap. To perform partial bootstrapping, a system compares the set of data keys from a source index at the data source with the set of data keys from a destination index at the destination database to identify a difference between the two sets of data keys. The system then deletes the data keys in the destination index that do not appear in the source index (as identified in the difference between the two sets of data keys).

Abstract: A method for dynamically controlling the alignment of multi-nested Objectives and Key Results (OKRs) is implemented via an application service provider server including a processor. The method includes executing, via a network, an enterprise application on a remote computing system and causing surfacing of a user interface on the display of the remote computing system during the execution of the enterprise application, where the user interface corresponds to a goal-setting feature of the enterprise application. The method also includes receiving, via the surfaced user interface, user input including an alignment permission policy for a multi-nested OKR of an enterprise, where the alignment permission policy defines a list of enterprise users who are allowed to align child OKR objects to a parent objective of the multi-nested OKR. The method further includes applying the alignment permission policy to the multi-nested OKR.

Abstract: An example computing device may include memory and one or more processors. The one or more processors may be configured to parallel entropy decode encoded video data from a received bitstream to generate entropy decoded data. The one or more processors may be configured to predict a motion vector based on the entropy decoded data. The one or more processors may be configured to decode a motion vector residual from the entropy decoded data. The one or more processors may be configured to add the motion vector residual and motion vector. The one or more processors may be configured to warp previous reconstructed video data with an overlapped block-based warp function using the motion vector to generate predicted current video data. The one or more processors may be configured to sum the predicted current video data with a residual block to generate current reconstructed video data.

Abstract: Techniques are described for performing transformation on video data. A transform circuit may receive M sample values of the video data from a pre-transform buffer, and process the M sample values with N computation units of the transform circuit to generate intermediate values. Processing the M sample values to generate the intermediate values includes feeding back temporary values from output of one or more of the N computation units to input of one or more of the N computation units. The transform circuit may store a first set of the intermediate values in a transpose buffer, and store a second set of the intermediate values in the pre-transform buffer that are to be later retrieved for storage in the transpose buffer.

Abstract: Methods, systems, and devices for motion analysis are described. Generally, the described techniques provide for computationally efficient and accurate motion analysis. A device may identify frames of a video frame sequence having a defined resolution. The device may downscale the frames to generate a plurality of downsampled images each having a resolution lower than the defined resolution. The device may generate a respective histogram vector for each pixel of each downsampled image and each pixel of the original frames. The device may determine a motion vector candidate based at least in part on the histogram vectors. The device may apply a filter to the motion vector candidates to determine a final motion vector and output an indication of motion between the frames of the video frame sequence based at least in part on the final motion vector for each pixel of the second frame.

Abstract: Techniques are described for performing transformation on video data. A transform circuit may receive M sample values of the video data from a pre-transform buffer, and process the M sample values with N computation units of the transform circuit to generate intermediate values. Processing the M sample values to generate the intermediate values includes feeding back temporary values from output of one or more of the N computation units to input of one or more of the N computation units. The transform circuit may store a first set of the intermediate values in a transpose buffer, and store a second set of the intermediate values in the pre-transform buffer that are to be later retrieved for storage in the transpose buffer.