Digital Commons Network^™

Autoregressive Summarization Of Trace Data Using Fine-Tuned Language Models, Jianan Tang, Nihar Dandekar, Saurabh Gupta

Defensive Publications Series

An autoregressive summarization process is proposed herein that can be used to assess trace data in an observability system, which provides a trade-off between storing state context and trace context. In some instances, the autoregressive summarization process may also be used to detect and prove differences between competitive offerings.

Powering Small Devices Using Micro Power Generation, Joel Richard Goergen, Tushara Ramesh, Emily Richter, Yuying Li

Defensive Publications Series

A micro power generation system is proposed herein that can be used to power a small device, such as a wireless access point (AP), to reduce data center power demands for low powered devices, thereby making such devices self-sufficient. The power generation system utilizes air duct flow, batteries, small generators with low starting resistance, and hybrid turbine blades. Tests indicate that such a system can consistently generate 5 watts or more.

Personalized Automatic Query Expansion Suggestions, Na

Defensive Publications Series

While autocomplete suggestions are a feature of search engines and other applications, these are typically limited to a few words and are one-shot. Interactions with a large language model (LLM) can benefit from rich, detailed prompts since the quality of responses provided by the LLM can improve with such prompts. However, such prompts are laborious to write. This disclosure describes techniques that leverage a large language model (LLM) to automatically suggest unstructured, natural-language query completions based on LLM reasoning. With user permission, appropriate user data can be accessed and used for the specific purpose of query completion. The techniques can …

Contract Lifecycle Management Within An Integrated Digital Workspace, N/A, N/A

Defensive Publications Series

Business agreements contain critical obligations that are typically extracted and tracked manually. This process is labor-intensive and prone to human error, which can lead to missed deadlines, breached terms, and potential liabilities. This disclosure describes a method for autonomously managing contractual obligations within a digital workspace. The method utilizes a large language model to analyze a contract, identify the contracting parties, and assign specific duties to the user's organization. Based on this party-specific assignment, actions such as scheduling calendar events, drafting payments, or populating organizational document templates are automatically orchestrated within the user's integrated suite of productivity applications. This approach …

System And Method For Explainable Deep Learning Fraud Detection Scores, David Sutton, Mark Anderson, Piotr Skalski, Yulia Rozanova

Defensive Publications Series

The present disclosure relates to a system for explaining deep learning fraud scores. The system comprises a deep learning model configured to process a sequence of raw transaction attributes and generate a fraud score. A concept definition framework includes a programmatically defined label set capturing semantic concepts associated with transactional behaviours. Concept activation vectors (CAVs) are established for each defined semantic concept based on historical transaction data. An integrated conceptual sensitivity (ICS) mechanism combines attributions from the deep learning model to the defined semantic concepts, providing an explanation for the generated fraud score. Additionally, a hierarchical explanation interface visualizes the …

Implementation Of Common Controls Framework Using Graph Database And Vector Database Technologies, David J. Driggers

Defensive Publications Series

This publication discloses methods, systems, and techniques for implementing a common controls framework using graph database and vector database technologies. Unlike traditional relational database implementations, these approaches leverage the inherent relationship modeling capabilities of graph databases and the semantic similarity search capabilities of vector databases to provide enhanced control mapping, gap analysis, and compliance assessment functionality. This disclosure establishes prior art for non-relational database implementations of security control frameworks.

Energy-Cost And Grid-Aware Scheduling Of Computational Tasks On Hardware Accelerators, N/A

Defensive Publications Series

The technology described in this paper relates to scheduling large-scale computational tasks, such as those related to large language models (LLMs), on specialized hardware accelerators. The technology utilizes a decision engine that dynamically evaluates external energy market data, utility grid status, and renewable generation forecasts concurrently with internal task characteristics, including service level objectives (SLOs). By employing a multi-factor cost function, the technology seeks to balance adherence to task deadlines against minimization of operational electricity costs and reduction of strain on the power grid. This allows for high-priority, low-latency requests to be processed immediately, while delay-tolerant batch requests are strategically …

Bias-Corrected Multi-Agent User Feedback Analysis System For Llm Performance Optimization, Saurabh Gupta, Andrea Morandi

Defensive Publications Series

Proposed herein is a novel bias-corrected multi-agent user feedback analysis system that automatically corrects for selection bias in sparse user feedback to provide accurate, unbiased quality assessments of Large Language Model (LLM) systems.

Novel Crystalline Form Of 4-[3¬ Chloro-4-(N’-Cyclopropylureido)Phenoxy]-7-Methoxyquinoline-6-Carboxamide Paratoluenesulfonate And Its Process For The Preparation, Srinivasan Thirumalai Rajan, Sagyam Rajeshwar Reddy, Kammari Balraju, Kamutam Sathish, Msn Laboratories Private Limited, R&D Center.

Defensive Publications Series

The present disclosure relates to a novel crystalline form of 4-[3­ chloro-4-(N’-cyclopropylureido)phenoxy]-7-methoxyquinoline-6-carboxamide paratoluene sulfonate which is represented by the following structural formula-1b, which is referred to as Lenvatinib paratoluene sulfonate.

Formula-1b

The present disclosure also relates to a process for the preparation of novel crystalline form of 4-[3­ chloro-4-(N’-cyclopropylureido)phenoxy]-7-methoxy quinoline-6-carboxamide paratoluenesulfonate of formula-1b.

Solid State Forms Of {(2s)-4-[7-(8-Chloronaphthalen-1-Yl)-2-{[(2s)-1-Methyl Pyrrolidin-2-Yl]-Methoxy}-5,6,7,8-Tetrahydropyrido[3,4-D] Pyrimidin-4-Yl]-1-(2-Fluoroprop-2-Enoyl) Piperazin-2-Yl} Acetonitrile And Their Processes For Preparation Thereof, Srinivasan Thirumalai Rajan, Sagyam Rajeshwar Reddy, Kammari Bal Raju, Abhijit Garai, Msn Laboratories Private Limited, R&D Center, Hyderabad, India.

Defensive Publications Series

Abstract:

The present disclosure relates to solid dispersions comprising {(2S)-4-[7-(8-chloronaphthalen-1-yl)-2-{[(2S)-1-methyl pyrrolidin-2-yl]-methoxy}-5,6,7,8-tetra hydropyrido[3,4-d] pyrimidin-4-yl]-1-(2-fluoroprop-2-enoyl) piperazin-2-yl} acetonitrile of formula-1 and one or more pharmaceutically acceptable excipients and processes for preparation thereof. The structure of formula-1 is shown as below:

Formula-1.

Novel Polymorphs Of Fezolinetant And Processes For Preparation Thereof, Srinivasan Thirumalai Rajan, Sagyam Rajeshwar Reddy, Kammari Balraju, Garai Abhijit, Msn Laboratories Private Limited, R&D Center, Hyderabad, India.

Defensive Publications Series

The present disclosure relates to a novel polymorphs of Fezolinetant of formula-1 which is represented by the following structural formula-1 and its process for the preparation.

Formula-1.

Providing Unified Interface For Rendering User Interface (Ui) Components And Application Programming Interface (Api) Documentation, Thirumalaa Srinivas , Visa, Summer Young , Visa, Miriam Salzer ,Visa, Murphy Basore , Visa

Defensive Publications Series

The present disclosure relates to a system and a method for dynamically fetching, filtering, and rendering one or more user interface (UI) component examples and application programming interface (API) documentation for multiple libraries and versions. The system enables users to select a library and version. The system fetches a corresponding metadata and code examples from a structured JavaScript Object Notation (JSON) file. The system provides device emulations, code/preview toggling, and API reference display in a unified interface.

Optimizer System And Method, Sayantika Sarkar, John H. Li, Shari Krikorian, Raymond B. Choi

Defensive Publications Series

The present disclosure provides a system and method for managing the payment of invoices using an optimizer system. The method comprises receiving, by an optimizer system, an accounts receivable file comprising a plurality of invoices from an accounts payable system, the accounts payable system configured to process invoices provided by a supplier to a buyer for payment; analyzing, by the optimizer system, the plurality of invoices in the accounts receivable file; determining, by the optimizer system, one or more invoices in the plurality of invoices to accelerate for payment; transmitting, by the optimizer system, a first message to the accounts …

Combined Power, Temperature, And Humidity Cycle Testing For Battery Reliability Assessment, Sandhya Chandrashekhar

Defensive Publications Series

Some battery reliability assessments may utilize separate, isolated tests for thermal, electrical, and environmental stress. These approaches may not fully capture complex failure modes that can arise from combined effects in certain operating conditions. A methodology is described for power temperature humidity cycling (PTHC) testing. The method can involve simultaneously subjecting a battery (which may be part of a device such as a smartphone, smart watch, or other wearable) to continuous and potentially asynchronous cycles of power (e.g., charge/discharge), temperature (for example, from -20°C to 60°C), and relative humidity (for example, from 10% to 90%). By combining these stressors, which …

Context-Aware Adjustment Of Noise Cancellation For Wearable Audio Devices, Shailesh Maheshwari, Vedant Maheshwari, Rayansh Maheshwari

Defensive Publications Series

Active noise cancellation in wearable audio devices, such as headphones or earbuds, can reduce a user's ability to hear certain environmental sounds. A system is disclosed that can implement context-driven audio awareness. The system may utilize one or more machine learning models to monitor a user's acoustic environment for specific, predefined sounds from a user-customizable list, such as sirens, doorbells, or a baby crying. Upon detection of a designated acoustic event, the system can automatically and temporarily modify the noise cancellation feature, for example, by disabling it or switching to an audio passthrough mode, allowing the external sound to become …

Novel Crystalline Form Of 6-[(6,7-Dimethoxyquinazolin-4-Yl)Oxy]-N,2- Dimethyl-1-Benzofuran-3-Carboxamide And Its Process For The Preparation, Msn Laboratories Private Limited, R&D Center; Srinivasan Thirumalai Rajan; Sagyam Rajeshwar Reddy; Gade Srinivas Reddy; Kammari Balraju; Bingi Hariprasad Reddy; Padigapaty Shiva Reddy And Matapathi Mahesh.

Defensive Publications Series

This application relates to a novel crystalline form of 6-[(6,7-dimethoxy quinazolin-4-yl)oxy]-N,2-dimethyl-1-benzofuran-3-carboxamide, also known as Fruquintinib. The disclosure also covers the process for preparing this new crystalline form.

Formula-1.

The present disclosure also relates to a process for the preparation of compound of formula-9 which is a key intermediate used in the preparation of Fruquintinib.

Formula-9

The present disclosure also relates to a process for the preparation of Fruquintinib.

Method And System For Alias Management Techniques, Suresh Kumar Tavvala, Wenhai Pan, Rajiv Sharma, Michelle Hou

Defensive Publications Series

A server computer receives, from a user device comprising an application comprising an SDK (software development kit), an onboard request, the application supported by an application server computer. The server computer provides an API key to the application in the user device. The server computer receives, from the SDK, an initialization request comprising the API key. A secure session is initiated between the server computer and the SDK in the user device. The server computer receives, from the SDK, a credential during the secure session and verifies the credential. The server computer stores, in an alias directory, the credential in …

A Housing For Multiple Pressure Sensors, Christian Ruwe, Thomas Heyen

Defensive Publications Series

The disclosed invention relates to a compact and modular pressure sensor housing capable of supporting multiple sensor configurations within a single device. Industrial applications often require both failsafe and non-failsafe pressure sensing systems, necessitating the use of one or more sensors such as gauge or differential pressure sensors based on safety and performance requirements. Traditionally, this has required separate housing designs for each configuration, increasing manufacturing complexity and cost. The proposed solution addresses this challenge by enabling the integration of up to four pressure sensors (two gauge and two differential) into a single housing through precise internal machining and strategically …

Prosodic Based Turn Detection Using The Audio Spectrogram Transformer, Adhwaith Menon

Defensive Publications Series

End of turn (EoT) detection is performed in a voice artificial intelligence (AI) solution to detect when a user is done talking. The currently available systems that perform EoT detection face issues such as detection of only the long pauses and added overhead cost from conversion of audio to text for analysis. To overcome the aforementioned issues, a system is proposed herein that uses a pre-trained Audio Spectrogram Transformer (AST) as a prosodic feature extractor, combined with a high-level transformer encoder that analyzes global prosodic structure. By leveraging this architecture, the system reliably predicts when a speaker has finished their …

Smart Device Hvac Health Evaluation With Large-Language Model Acoustic Signature Analysis, D Shin

Defensive Publications Series

This publication describes systems and techniques for evaluating the operational status or “health” of a heating-ventilation-air-conditioning (HVAC) system by analyzing sound generated by the HVAC system. The sound may be monitored by a microphone incorporated in a smart thermostat or other smart home device and automatically evaluated using an HVAC Large Language-Type Model (HVAC-LLM) including acoustic signatures of properly and/or improperly operating HVAC systems. For example, a system includes one or more microphones integrated in a smart thermostat and/or one or more other smart home devices that act as passive sensors to capture sound made by the HVAC system. Initialization …

Event-Triggered Audio Mode Management For Hearable Devices, Shailesh Maheshwari, Vedant Maheshwari, Rayansh Maheshwari

Defensive Publications Series

This document describes a method for managing audio modes in a hearable device. A user operating a device in an ambient sound passthrough mode may have difficulty hearing audio events, such as notifications or digital assistant responses, over background noise. A system can detect certain triggers, such as the receipt of a notification or the start of a user's interaction with a digital assistant. Upon detecting a trigger, the hearable device can temporarily switch from an ambient sound mode to a noise cancellation mode. After the audio event concludes, for example, after the notification sound or the assistant's response is …

Structured Layered Synthesis For Generating Coherent And Scale-Consistent Text, Saurabh Gupta, Andrea Morandi

Defensive Publications Series

Although Large Language Models (LLMs) excel at generating fluent text at a local level, LLMs struggle to consistently maintain the structure and coherence of generated text over long documents and across different levels of detail. Techniques proposed herein introduce a system and method for enhancing text generation through a structured layered synthesis approach.

Situational Awareness In Virtual Reality By Providing A Wide Field Of View, Alexander Graehl Williams

Defensive Publications Series

Virtual reality (VR) headsets typically limit a user's field of view, restricting situational awareness to what is directly in front of them. This makes it difficult to monitor the surrounding environment for events or motion occurring outside the primary viewing area. This disclosure describes a method for conditionally displaying a panoramic, 360-degree view within a VR headset. The panoramic view can be triggered by events, such as the detection of motion in the user's periphery or rear. Upon detection, the standard forward-facing view is temporarily replaced or augmented with a compressed 360-degree visualization of the environment. This technique enhances the …

Stateful News Ranking To Prioritize Novel Content For Repeat Queries, Deep Shah, Aditya Bhandari

Defensive Publications Series

A method for stateful ranking of search results may be used to prioritize novel content when a user repeats a query, for example, for a developing news story. A system can establish a "context timestamp" based on a user's most recent prior query on a topic. The system may then identify "context results," which can be items published before this timestamp, and "latest results," which can be items published after it. To differentiate new information, the system may computationally compare the latest results to the context results, for instance, by generating vector embeddings and identifying latest results with a similarity …

Generating Context-Aware Code-Switched Text Using Probabilistic Modeling And Generative Refinement, Mithun Kumar S R

Defensive Publications Series

Artificial intelligence models can experience difficulty understanding and generating code-switched text, and some existing synthetic data generation methods can produce grammatically incorrect or unnatural-sounding results. This disclosure describes a multi-stage system for generating code-switched text from monolingual corpora. The method can employ a probabilistic constraint model to analyze a source sentence and identify syntactically and semantically appropriate points for a language switch. Based on these identified points, a candidate generation engine can create one or more raw code-switched sentences. A large generative model may then refine these raw candidates to address potential grammatical errors at the language junction and improve …

System And Method For Topologically Protected Quantum Computing Based On Majorana Fermions In Superconducting Nanostructures With Integrated Cryogenic Control, Lorenzo Lbn Bernardini, Origines Technologies

Defensive Publications Series

The present invention is located in the field of topologically protected quantum computing, with particular reference to the implementation of qubits based on Majorana fermions in nanoscopic superconducting structures. This technological field represents an innovative convergence between condensed matter

in physics, quantum computing and superconducting materials engineering. The system object of the present invention operates in a technological domain where quantum hardware based on Majorana fermions is integrated with sophisticated topological protection mechanisms of quantum states. This integration is achieved through the implementation of qubits that exploit topologically protected states within nanoscopic superconducting structures, allowing a coherent manipulation of non-abelian …

Cooling-Aware Proactive Load Shaping In Data Centers, N/A

Defensive Publications Series

This disclosure describes techniques for proactively forecasting data center cooling capacity and cooling load(s) in order to anticipate and mitigate potential shortfalls in cooling capacity. Per techniques of this disclosure, a dynamic, forward-looking model of the thermal supply and demand for the data center is generated. The effective cooling capacity available at a data center for future periods is determined based on multi-day weather forecasts, equipment metadata, and historical power consumption patterns. Concurrently, the expected cooling load is estimated based on the computational workloads. When a potential risk is identified, mitigating actions are initiated. Such actions can include sending proactive …

On-Device Large Language Model Inference With Unlimited Context Size, Xusong Wang, Miao Wang

Defensive Publications Series

On-device large language model (LLM) pipelines that employ hardware accelerators frequently encounter limitations regarding the maximum context size that can be processed on-device. This constraint arises because neural network accelerators are designed to favor the execution of statically-shaped graphs, whereas varying context sizes necessitate dynamic dimensions within the attention layer of a neural network. Conventional solutions present challenges such as computational inefficiency and lack of scalability. This disclosure addresses the issue of context size limitation when performing inference using an on-device LLM by partitioning the attention graph into multiple reusable, statically-shaped graphs that are mathematically equivalent to the original computation. …

Automatically Generating High Value Tests For User Journeys Of An Application, Siva Velusamy

Defensive Publications Series

Software development includes writing tests and executing the tests to ensure that the software application works as intended. However, manually written tests are expensive, time-consuming, and can often be incomplete and/or may not be available. This disclosure describes the use of a large language model (LLM) to automatically generate test scenarios for the most common paths that real users take when using an application. Data regarding such common paths is available for many applications since application owners have a need to view analytics on which portions of their application are visited and the user engagement with different portions of the …

Method To Run Interactive Or Multistep Critical Patches Remotely For Computing Systems With Out-Of-Band Or In-Band Computing Resources, Hp Inc

Defensive Publications Series

This disclosure concerns remote administration of computing devices via keyboardvideomouse (KVM) consoles provided by out-of-band management hardware or in-band remote desktop software. The focus is secure, automated execution of interactive or multistep remediation at operating system, firmware, or bootloader layers where user input and timing govern success.