Abstract: A method may include receiving a primary document including one or more selection boxes, one or more text lines, and one or more annotations. The method may include determining, a class based on the annotations. The method may include identifying the one or more selection boxes and one or more text lines of the primary document. The method may include generating a graph representing the one or more selection boxes and the one or more text lines. The method may include mapping each of the one or more selection boxes to a respective text line of the one or more text lines of the graph based at least in part on one or more characteristics associated with the selection boxes. The method may include generating a key-value pair associated with each of the one or more text lines and generating a document model of the primary document.

Abstract: Systems, computer program products, and methods are described herein for resource transfer mode evaluation in distributed network using semi-dynamic tokenized graph node processing. The present disclosure is configured to receive resource data from one or more resource transfer channels; extract metadata from the resource data and determine one or more resource transfer processing requests; generate a dynamic hash value for the one or more resource transfer processing requests; tokenize the dynamic hash value to generate a semi-dynamic token; select a resource gateway and a resource mode for the one or more resource transfer processing requests; generate a key value pair for the selected resource gateway and the resource mode; tokenize the key value pair and store the tokenized key value pair on a distributed ledger; and flag one or more non-selected resource gateways and resource nodes.

Abstract: Techniques for implementing out of distribution element detection for information extraction are disclosed. In an example method, a cloud service provider (“CSP”) recognizes one or more elements in a document using an optical character recognition (“OCR”) technique. The CSP classifies each element of the one or more elements as belonging to (1) an in-distribution (“in-D") class of one or more in-D classes or (2) an out-of-distribution (“OO-D”) class using a machine learning (ML) model, in which an element is classified as belonging to the OO-D class when it is not classified as belonging to any of the one or more in-D classes.

Abstract: Systems and methods are disclosed for optimizing webpage loading times. One method may include receiving one or more content slots of a webpage, classifying the one or more content slots, batching the one or more content slots based on the classifying of the one or more content slots, customizing one or more bid requests corresponding to each batch of the one or more content slots, transmitting the one or more customized bid requests to an auction manager, receiving, from the auction manager, one or more winning bids for each of the one or more bid requests, and loading the one or more content slots of the webpage with one or more electronic content items corresponding to the one or more winning bids.

Abstract: A graphical user interface (GUI) of a computer system is described herein. The GUI comprises a display (e.g., screen) comprising a first sector and a second sector that may be used to assist a user to perform a surgical intervention, e.g., a cardiac ablation procedure using a catheter having a balloon with electrodes disposed thereon. The first sector may include a menu comprising dialog boxes. The second sector may comprise an image of the balloon and representations of the electrodes. The GUI may also include a processor. The processor may be configured to, e.g., receive inputs from the dialog boxes, change appearances of the dialog boxes, receive inputs from the electrode representations, change appearances of the electrode representations, control irrigation to the balloon, and control power to the electrodes.

Abstract: Artificial intelligence chatbot performance is improved through automated conversation quality analysis and training example generation. Chatbot conversations are processed via dual pathways, a sentiment analysis model measuring user satisfaction and an alignment analysis model evaluating query-answer semantic correspondence. These analyses produce alignment and sentiment scores that are combined into a sentiment and alignment deviation composite score called a “SADI score.” The SADI score quantifies response accuracy and satisfaction changes across conversation segments, enabling identification of problematic patterns. When SADI scores indicate significant deviations from performance thresholds, the new training examples are generated from the problematic conversations. The generated examples focus on segments showing sentiment or alignment degradation. A target machine learning model receives parameter updates using these examples, creating a feedback loop where user interactions drive model refinement.

Abstract: Certain aspects of the disclosure provide methods, systems, and apparatuses for providing taxonomy-boosted hierarchical hidden Markov modeling to enhance recommendation inferences and machine learning model training. A hierarchical structure of enterprise knowledge is constructed to provide searching knowledge in a desired granularity. A hidden state space may be defined for a hierarchical hidden Markov model, where each hidden state in the hidden state space is associated with a node in the hierarchical structure and an intent category obtained from observing the actions, events, and behavior of the user. Training a classification model may be provided to predict a probability of each hidden state in the hidden state space based on a user input. Training the hierarchical hidden Markov model may be provided based on a training data set labeled for each hidden state in a hidden state space.

Abstract: Techniques for hybrid retrieval using a knowledge base (KB) system for responsive multi-turn conversations are disclosed. The system evaluates user queries by generating confidence scores from comparisons against a set of predefined intents. When a confidence score falls within a high-confidence range, the system produces a predefined intent-based response. When a confidence score falls within a low-confidence range, the system retrieves a relevant KB document and generates a response by applying a language model (LM) to the query and the retrieved document.

Abstract: Techniques for configuring a machine learning (ML) pipeline are disclosed. A system generates a region comprehension index (RCI) value that quantifies, for a visual dataset supplied to a computer vision model (CVM), a bias toward global or local reasoning. The system determines a first performance metric at least by applying the CVM to a whole image included in the visual dataset and a second performance metric at least by applying the CVM to patches of the whole image. The system determines the RCI value as a function of at least the first performance metric and the second performance metric. If the RCI value satisfies a criterion for deploying the CVM to an ML pipeline, the system selectively deploys the CVM to the ML pipeline.

Abstract: Techniques for configuring a machine learning (ML) pipeline are disclosed. One or more embodiments generate a patch context robustness index (PCRI) value that quantifies, for computer vision model (CVM), sensitivity of the CVM to contextual noise, at least by: determining a first performance metric at least by applying the CVM to a whole image; determining a second performance metric at least by applying the CVM to patches of the whole image; and determining the PCRI value as a function of at least the first performance metric and the second performance metric. One or more embodiments deploy the CVM to an ML pipeline based at least on the PCRI value.

Abstract: In some aspects, the techniques described herein relate to a method including: detecting, by a processor, an interstitial break during playback of a primary media asset; identifying, by the processor, an interstitial associated with the interstitial break, the interstitial associated with a type; instantiating, by the processor, a renderer based on the type of the interstitial; and executing, by the processor, the interstitial using the renderer.

Abstract: Techniques are disclosed herein for generating dynamic vocabularies for conditioning a language model. A dynamic vocabulary is constructed from an input prompt, database schema information for a database to be queried, and programming language information for a programming language to be used for querying the database to condition the language model to predict an output statement in the programming language. The dynamic vocabulary can be included in prompt information that is provided to the language model. The number of tokens in the dynamic vocabulary can be different than a number of tokens included in a vocabulary of the language model. By utilizing a dynamic vocabulary, the language model can be conditioned to predict tokens for the output statement that are contextually consistent with the tokens included the dynamic vocabulary.

Abstract: A method and apparatus for verifying object classification includes comparing detected object-components with trained component-based descriptors and similarity measures to generate object classification verification data. The object classification verification data represents object classification confidence or misclassification errors to weight uncertainty for perception-based decision making. A sequence of similarity measures associated with a current frame and prior frames within a reference observation time constraint are compared to a reference temporal similarity measure boundary to generate the object classification verification data.

Abstract: Techniques for extracting information from unstructured documents that enable an ML model to be trained such that the model can accurately distinguish in-distribution (“in-D”) elements and out-of-distribution (“OO-D”) elements within an unstructured document. Novel training techniques are used that train an ML model using a combination of a regular training dataset and an enhanced augmented training dataset. The regular training dataset is used to train an ML model to identify in-D elements, i.e., to classify an element extracted from a document as belonging to one of the in-D classes contained in the regular training dataset. The augmented training dataset, which is generated based upon the regular training dataset may contain one or more augmented elements which are used to train the model to identify OO-D elements, i.e., to classify an augmented element extracted from a document as belonging to an OO-D class instead of to an in-D class.

Abstract: The present disclosure relates to a custom framework for fine-grained human activity recognition. One or more input videos may be accessed, where the one or more input videos comprise one or more frames depicting one or more actors and one or more objects. A plurality of object-pose interaction graphs may be generated for individual frames from the one or more input videos based at least in part on one or more objects of interest from the one or more objects and on one or more joint keypoints of the one or more actors. A first graph neural network may be trained based at least in part on the plurality of object-pose interaction graphs to identify spatial information for the one or more actors, the one or more objects of interest, and one or more interactions between the one or more actors and the one or more objects of interest.

Abstract: Techniques for evaluating a language model for multilingual performance are disclosed. A framework generates prompts having components in a first language and in a different language. The different components include at least: a question, instruction, context, or instructed response language. The prompts test an ability of an LLM to respond in a specified language, an ability of the LLM to reason correctly, and/or an ability of the LLM to respond consistently. The framework calculates language token scores and fine-tuning performance scores. The framework generates a total score by weighting a token score, fine-tuning performance score, and/or one or more task scores. The scores are usable as a criteria for deployment of an LLM, for selecting training data or techniques, or for determining training weights used during tuning of an LLM.

Abstract: Techniques for multilingual batch alignment training and/or fine-tuning for language models are included herein. The system generates and/or accesses data points from batches of data having semantically equivalent content across languages and/or semantically equivalent prompts across languages. The prompts include queries related to a topic or a predictive language task. The data points include one or more sets of resources that are semantic equivalents in different languages. Data points for which a resource enables a correct response are used as positive training data points for reinforcement learning. Data points for which a resource does not enable a correct response are considered as negative training data points for reinforcement learning. The data points are used for alignment training and/or other reinforcement learning for generative language models.

Abstract: A method may include receiving a primary document including one or more selection boxes, one or more text lines, and one or more annotations. The method may include determining, a class based on the annotations. The method may include identifying the one or more selection boxes and one or more text lines of the primary document. The method may include generating a graph representing the one or more selection boxes and the one or more text lines. The method may include mapping each of the one or more selection boxes to a respective text line of the one or more text lines of the graph based at least in part on one or more characteristics associated with the selection boxes. The method may include generating a key-value pair associated with each of the one or more text lines and generating a document model of the primary document.

Abstract: Techniques for evaluating a language model are disclosed herein. One or more agents and/or one or more language models are used to generate responses to queries in conversations. For example, during chat session conversations, the agent and/or model used to generate responses to queries in the conversation is sometimes changed. Various metrics are collected for the conversations for which a prompt, agent or model change occurs and/or the conversations for which a change to the prompt, agent and/or model does not occur. Metrics such as success rate, escalation rate, query repetition, direct feedback, or other feedback is collected and/or aggregated for groups of users and/or topics of conversations. One or more models used to generate responses in the conversations are evaluated and/or selected based on the metrics.

Abstract: In some implementations, techniques described herein may include identifying text in a visually rich document and determining a sequence for the identified text. The techniques may include selecting a language model based at least in part on the identified text and the determined sequence. Moreover, the techniques may include assigning each word of the identified text to a respective token to generate textual features corresponding to the identified text. The techniques may include extracting visual features corresponding to the identified text. The techniques may include determining positional features for each word of the identified text. The techniques may include generating a graph representing the visually rich document, each node in the graph representing each of the visual features, textual features, and positional features of a respective word of the identified text. The techniques may include training a classifier on the graph to classify each respective word of the identified text.