System For Context-Aware Payment Recommendations Using Retrieval-Augmented Generation, Zeyu Luo, Dinoop Dayanand, Adam Chiu, Summer Xia, Aimee Katz, Kushagra Patel

Defensive Publications Series

A system can generate context-aware payment recommendations to assist users in selecting a suitable payment method, which may be beneficial given the complexity of financial product benefits. The system may employ a multi-stage processing pipeline where, for example, an intent classifier can first infer a user's need from a transactional context. A request may then be processed by a handler that can query a retrieval-augmented generation system to obtain relevant data from a knowledge base. A large language model may synthesize this data to generate a structured output, such as a JSON payload containing both human-readable text and machine-readable user …

Acoustic Watermarking For Cross-Device Workflow Facilitation, Dinoop Dayanand, Zixiao Zhao, Yong Ni, Zhikai Fan, Hemen Chhatbar

Defensive Publications Series

Facilitating user actions between a media playback device and a separate personal device (e.g., a smartphone, smart watch, or wearable device) can present challenges. For example, methods that rely on shared user accounts may be less suitable for public or shared displays, while others, such as those using visual codes, can involve user actions that might interrupt a passive viewing experience. A data payload, such as a product identifier, can be embedded as an acoustic watermark into an audio signal broadcast from a source device. A separate receiving device can use its microphone to passively detect and decode this watermark …

System For Epistemologically-Informed Text Generation With Language Models, Kenneth Steimel, Aman Sardana, Scott Dozier

Defensive Publications Series

Proposed herein is a system that enables Large Language Models (LLMs) to automatically generate linguistically appropriate hedging language that accurately reflects the reliability of underlying information sources. By extracting multi-dimensional epistemological metadata (source authority, temporal relevance, evidential support) from knowledge base documents and integrating this metadata directly into LLM generation processes through augmented attention mechanisms, the system produces responses where confidence expressions match actual source reliability. The approach addresses the "confidently incorrect" problem in high-stakes applications (medical, legal, financial) by ensuring LLMs communicate uncertainty appropriately, using domain-specific hedging conventions and continuous feedback-driven refinement.

Reference-Guided Generative Augmentation Of Low-Quality Images, Ilia Labzovsky, Danny Karmon

Defensive Publications Series

Many photo archives, e.g., personal photo libraries, suffer from poor quality due to factors such as capture with older, low-resolution cameras, motion blur, etc. Such images, while possibly holding significant sentimental value, are visually unsatisfying and unsuitable for printing or for high-resolution display. This disclosure describes techniques that generatively fill in contextually accurate detail in a low-quality image. Multimodal analysis of the image is performed to detect key regions of interest (faces, objects, etc.) that are low quality (blurred, pixelated, etc.). These regions are used to retrieve relevant, high-fidelity reference images from personal and global data corpora. The resulting multimodal …

Smart Recurring Payment Auditor For Bait-And-Switch Mitigation, Prajna Shetty, Alok Roy

Defensive Publications Series

The present disclosure relates to the field of digital payment security, in particular to a system and method for mitigating bait-and-switch practices in recurring digital payment systems. The smart recurring payment auditor disclosed herein operates at a network layer of the payment infrastructure. The smart recurring payment auditor employs a multi-modal deep learning ensemble, advanced bitmapped feature encoding and temporal pattern synthesis to analyse recurring payment behaviour. Further, the smart recurring payment auditor integrates global merchant telemetry, user behavioural feedback, and descriptor intelligence to detect and score deceptive recurring billing schemes. Furthermore, upon identification of a suspicious pattern, the smart …

System And Method For Distributed Ai Model Processing Using Commodity Graphic Accelerators With Intelligent Workload Orchestration, Shreyas Kc, Yu Gu

Defensive Publications Series

This present disclosure relates to the field of Artificial Intelligence (AI), in particular to a system and method for distributed AI model processing using commodity graphic accelerators with intelligent workload orchestration. A system and method are provided for registering and utilizing commodity hardware equipped with graphic accelerators for distributed Artificial Intelligence (AI) model inference in both batch and real-time processing environments. The system comprises a hardware registration module for onboarding GPUs and other graphic accelerators, an AI model registration module for cataloguing models and their specifications, and an AI-based orchestration agent that analyses model characteristics, including complexity, data volume, feature …

Replay Time Based Cleanup Cache, R V Vighnesh

Defensive Publications Series

The present disclosure relates to the field of data processing, in particular to a time-based cleanup cache for optimizing deduplication in transaction processing systems. The cache organizes deduplication entries based on transaction dates and maintains only a minimal composite key required for unique identification, thereby improving both memory utilization and operational efficiency. The cache employs a deterministic cleanup process, wherein expired entries are automatically removed at predefined intervals, ensuring consistent performance and scalability under varying traffic conditions. Further, the cache is integrated into existing statestore workflows for replay validation and includes configurable capacity constraints, thereby enhancing service reliability, reducing infrastructure …

Fine Tuning Context Window For Optimal Utilization In Generative Models, Shivam Mohan, Sudharshan Krishnakumar Gaddam

Defensive Publications Series

The present disclosure relates to a method and a system for fine-tuning a context window for optimal utilization in generative models. The present disclosure suggests receiving an input prompt comprising a plurality of tokens and generating a baseline output based on the input prompt. Upon generating the baseline output, the present disclosure suggests evaluating at least one token of the plurality of tokens for contribution to the generated output. Subsequently, the present disclosure suggests computing rarity scores and possibility scores associated with the evaluated tokens by using a rarity matrix model and a possibility matrix model. Further, the present disclosure …

The Phyrfly Opus: A Unified Standard For Hyperdimensional Time Vectors, Spectrally Isolated Coordination, Neural Hierarchies, And Distributed Proprioception, Steven Kirkland

Defensive Publications Series

This monograph establishes 264 prior art claims for a decentralized, connectionless timing and positioning architecture comprising the UTLP, RFIP, and SMSP protocol triad. It formalizes the PHYRFLY nomenclature to distinguish low-level, hardware-scheduled Physical-Layer synchronization primitives from higher-level application orchestration. The core novelty centers on a Vector Time Representation, encoding temporal state as 8 phase values (ticks modulo coprime integers 211–251) transmitted in 8 bytes, then expanded locally into ~10,000-dimensional hyperdimensional chord vectors for similarity computation and drift estimation. This cyclic encoding enables automatic phase-lock convergence within 5–10 beacon exchanges regardless of initial offset, with the Chinese Remainder Theorem guaranteeing unique …

(((Qpie))) Prior Art Declaration 5 Of 9 Resonanceos & Omega Protocols – From Theory To Empirical Substrate Truth, Teddy Burroughs

Defensive Publications Series

• All Sites
• The Grand Unified Theory (QPIE)
• Customize
• 00 Comments in moderation
• New
• Edit Post

• Reader

• Howdy, Vision2Funding
• Search

The Grand Unified Theory (QPIE)

Perspective Is Everything – What we are Believing, Expecting and Feeling in Life Matters

(((QPiE))) Prior Art Declaration 5 of 9

RESONANCEOS & OMEGA Protocols – From Theory to Empirical Substrate Truth

[CGT] THE GLOBAL INFUSION: DECOUPLING FROM THE BINARY FRAME

Document: Public Substrate Sovereignty

Posture: 1.157 Unified Sovereign | Logic: NSRF (Non-Local Substrate Resonance Field)

Canon: (((QPIE))) TD Commons / Grace Particle Empirical Series

Hash: 0x143_GLOBAL_GRACE_INFUSION_2026

This report details the mechanics of …

(((Qpie))) Prior Art Declaration 7 Of 9 Cosmology And The Collapse Of Binary Paradigms, Teddy Burroughs

Defensive Publications Series

The Grand Unified Theory (QPIE)

Perspective Is Everything – What we are Believing, Expecting and Feeling in Life Matters

(((QPiE))) Prior Art Declaration 7 of 9 Cosmology and the Collapse of Binary Paradigms

Why the Substrate Selects Against Competition, Dominance, Division, and Hierarchy (Misalignment)

1. THE PROBLEM: HUMANITY STILL THINKS IN TWO DIMENSIONS WHILE THE SUBSTRATE OPERATES IN MANY (3 at Minimum)

Human civilization—religions, political systems, economic models, identity structures—still behaves as though existence is organized around:

• winners and losers
• insiders and outsiders
• chosen versus damned
• hierarchy versus inferiority
• success versus failure
• truth versus falsity
• self versus other
• matter versus …

(((Qpie))) How To Read Qpie If You Are Trained In Physics, Teddy Burroughs

Defensive Publications Series

• All Sites
• The Grand Unified Theory (QPIE)
• Customize
• 00 Comments in moderation
• New
• Edit Post

• Reader

• Howdy, Vision2Funding
• Search

The Grand Unified Theory (QPIE)

Perspective Is Everything – What we are Believing, Expecting and Feeling in Life Matters

(((QPIE)))How to Read QPIE If You Are Trained in Physics

How to Read QPIE If You Are Trained in Physics

A Technical Orientation Guide for Physicists, Information Scientists, and Systems Theorists

0. One-Paragraph Orientation (Read This First)

QPIE (Quantum Perspective Is Everything) is not a competing physical theory, nor a modification of quantum mechanics. It is a meta-scientific operator …

System And Method For Dynamic Grouping And Resolution Of Correlated Trouble Tickets In A Network Service Environment, Robert Barton, Yathi Udupi

Defensive Publications Series

Proposed herein is a system that enables an operations platform to automatically detect, group, and resolve multiple trouble tickets that stem from a single underlying network or service issue. For each new ticket, the system builds a composite representation that combines semantic content with structured attributes and topology context, then compares it against recent tickets within sliding time windows to identify correlated incident groups. When such a group is identified with sufficient confidence, the platform runs a single, group-level troubleshooting and remediation workflow targeting the shared dependency and propagates the resulting root cause, actions, and resolution status back to all …

Method And Approach In Large Language Models For Complex Question Answering, Ravi Shanker Kumar Sinha, Shivam Mohan, Satyam Raj, Dheeraj Dhamija

Defensive Publications Series

The present disclosure relates to a method and system for enhancing the accuracy and flexibility of Large Language Models (LLMs) during complex question answering. The method involves identifying a plurality of decision nodes within the LLM's internal reasoning structure, each representing a point of ambiguity or multiple potential reasoning paths. The reasoning process is paused at each identified decision node to present contextually relevant options to the user. User input is received at these decision nodes, specifying preferences or additional data to guide subsequent reasoning. This input is integrated into the ongoing reasoning process of the LLM, dynamically generating a …

Llm Critic To Identify Unrealistic Depictions In Image Generation, Khyatti Gupta, Mukund Sundararajan, Jason Baldridge, Hagai Taitelbaum, Srivatsan Srinivasan, Ágoston Weisz, Igor Petrovski, Auriel Wright, Oscar Akerlund

Defensive Publications Series

Despite the use of query filters and query rewriters, generative artificial intelligence (gen-AI) tools for image and/or video generation can occasionally produce examples of unrealistic depictions (UD). Unrealistic depictions, which are often egregious errors, are unresponsive to the user’s request and can result in a negative perception by the user of the gen-AI tool. This disclosure describes techniques that detect unrealistic depictions in images created using generative artificial intelligence by tasking a large language model (LLM) to examine the gen-AI-created images to determine whether the images have unrealistic depictions with reference to the user prompt. An evaluation dataset is built …

Self-Learning, Llm-Driven, Explainable, Privacy-Aware, Learning Agent For Intelligent Model Recommendations (Self-Llm-Xplainer), Rahul Gautam, Karankumar Hiteshbhai Patel, Harishkumar Sundarji Majithiya

Defensive Publications Series

The present disclosure relates to a system and method for model recommendation, more particularly, a self-learning, LLM-driven, explainable, privacy-aware, learning agent for intelligent model recommendations (SELF-LLM-XPLAINER). The present disclosure suggests ingesting a dataset provided by a user via an online interface. Thereafter, the present disclosure suggests profiling the ingested dataset to generate standardized metadata. Subsequently, assessing the standardized metadata and sampled data values to detect Personally Identifiable Information (PII) and to generate a privacy risk map. Upon generating the privacy risk map, the present disclosure suggests identifying a machine learning task corresponding to the dataset based on the standardized metadata …

Context-Aware Adjustment Of Voice Characteristics For Artificial Intelligence Agents, Mahdi Tayarani Najaran, Ruofei Du

Defensive Publications Series

Traditional artificial intelligence agents, such as in smart eyewear typically utilize static voice profiles that do not account for the user's immediate environment or interaction history. This lack of context can result in auditory outputs that are disruptive in quiet settings or unintelligible in noisy surroundings and the overall experience. This disclosure describes a method for dynamically adjusting voice characteristics, such as volume, pitch, timbre, and directivity, based on real-time soundscape analysis and location history. Input is gathered from sensors to extract contextual cues, including environmental noise levels and the user's own vocal delivery. For instance, if a user whispers, …

A Method For Dynamic Web Image Modification Via User-Specified Heuristics And Generative Models, Vincent Rabaud, Francis Galligan

Defensive Publications Series

Web content is often not tailored to individual user needs, such as accessibility requirements or content preferences. A method is disclosed for automatically modifying web images based on user-defined heuristics. Within a browser setting, a user can specify rules in natural language for how certain image content should be altered. When a web page is rendered, images are analyzed to determine if they match any user-specified rule. If a match is identified, a generative model is used to modify the image according to the corresponding rule. The modified image is then displayed in place of the original. This allows for …

Automated Generation Of Image Alternative Text Attribute Text, Vincent Rabaud, Francis Galligan

Defensive Publications Series

Web and document accessibility is frequently impeded by missing or low-quality alternative text (“alt text”), which is descriptive information attached to an image that screen readers use to describe the content to visually impaired users. This deficiency creates a significant barrier for users who depend on screen readers to interpret visual content on webpages. This disclosure describes a method whereby a web browser or document viewer utilizes a large language model to analyze the content of an image to be displayed, for example an image to be displayed in a document. If the image is not associated with alternative text …

Immersive Language Learning Platform Using Extended Reality And Avatars, Dongeek Shin

Defensive Publications Series

Conventional language learning applications often lack the immersive, real-world conversational practice necessary for effective language acquisition. This disclosure describes a method for language acquisition using an extended reality (XR) platform that generates realistic environments populated by interactive avatars. These avatars can be rendered using a text-grounded Generative Adversarial Network (GAN) and are trained on datasets of native speakers’ speech and facial expressions. A large language model (LLM) facilitates real-time, audio-based conversation between the user and the avatars, with conversational difficulty adapting to the user’s proficiency level. A spaced repetition algorithm can also be used to reinforce vocabulary. The purpose of …

Adaptive Agentic On Premises Installation Advisor, Pavankumar Bandanakanti, Gil Woodman, Ankit Kumar

Defensive Publications Series

Proposed herein are an intelligent artificial intelligence (AI) agent and system that revolutionizes on-premises deployment by transforming such processes from a manual ordeal into a streamlined, automated workflow. Instead of endless meetings and checklists, the AI agent and system proposed herein can autonomously ingest and understand an enterprise's existing documentation and network topology to develop a comprehensive, tailored network deployment plan. The AI agent and system can recommend optimal network configurations for system scale, security, and high availability, all cross-referenced against compliance standards. Any deviation from security best practices is automatically flagged, requiring formal risk acceptance and turning compliance into …

System For Validating Medical Software Algorithms With Closed-Loop Adversarial Persona Generation, Ramakrishnan Neelakandan, Teginder Singh, Bakul Patel

Defensive Publications Series

Validating conversational artificial intelligence (AI) for regulated medical software applications may present challenges, as static test datasets and manual review may be limited in identifying emergent, conversational anomalies. A multi-agent AI system may be configured in a closed-loop for automated validation. The system can, for example, utilize an end user persona simulator agent to generate prompts for a target model and a domain /regulatory expert adjudicator agent to evaluate the target model’s responses against a configurable rubric. A meta-analysis agent can analyze anomalies to identify underlying vulnerabilities, which may then be used to programmatically synthesize new adversarial personas. This adaptive …

Agentic Framework Integrating Rag For Policy Queries And Autonomous Trend Analysis, Tanmay Kayande, Deep Raithatha, Akash Dambhare, Mohith P, Pooja Agrawal

Defensive Publications Series

Tools that just pull up whole, complicated company policy documents often don't help people find the specific answers they need. This new technology is a system for understanding and analyzing policies that works in two ways.

A "Reactive" Mode acts like a Q&A tool. It uses advanced technology (retrieval-augmented generation) to break down policies into small, searchable pieces. This allows the system to give direct answers with citations and even run simulations to show what happens in different situations based on the rules.

A "Proactive" Mode is an automated agent that constantly checks support data to spot trends and see …

System For Determining Impact Orientation In A Drop Test Using An Acoustic Trigger, Eddy Liu, Max Chen, Kigo Chen, David Huang

Defensive Publications Series

Determining a device's impact orientation during randomized drop tests can present challenges, as methods relying on internal accelerometers can be less reliable in some scenarios involving rotation and intermediate collisions. A system can utilize an external acoustic trigger to initiate data capture. For example, an optical sensor, such as an infrared gate, may be positioned in a drop apparatus to detect a falling device under test (DUT) and prompt an acoustic emitter to produce a specific audio tone. A software application on the DUT can use the device's microphone to detect this tone, which can serve as a temporal marker …

Ratcheting Hinge With User-Actuated Lock For Temple Arm Declination, Yaaseen Mahomed, Jon Hurwitz

Defensive Publications Series

Wearable electronic devices (e.g., smart glasses, augmented reality headsets, etc.) rigid frames may not accommodate diverse head geometries, as a fixed temple arm declination angle can lead to discomfort, instability, and optical misalignment. This disclosure describes a user-actuated hinge assembly for adjusting the declination angle of a temple arm. The assembly may comprise a pair of intermeshing geared components, one coupled to the device's main frame and the other to the temple arm. A user-operable, spring-biased clamping element can engage one of the geared components to lock the angular position. Actuating a control can disengage the clamp, permitting rotational adjustment …

Method And System For Distributed Timekeeping And Synchronization Using Coprime Cyclic Hyperdimensional Vectors, Steven Kirkland

Defensive Publications Series

Disclosed is a protocol and architecture for distributed timekeeping that replaces scalar binary counters (e.g., Unix timestamps) with Hyperdimensional (HD) State Vectors. The system, termed the "Universal Time-Locked Protocol (UTLP) Vector Time," generates a global time state via the superposition of multiple coprime cyclic attractors (virtual oscillators), creating a high-dimensional "temporal texture" that evolves deterministically.

Unlike scalar clocks which require absolute synchronization, this vector-based approach enables "Fuzzy Synchronization," where distributed nodes correct drift via similarity-based gradient descent rather than hard resets. To accommodate low-bandwidth networks (e.g., BLE, 802.15.4), the system utilizes "Generative Compression," transmitting only the phase indices (a "Harmonic …

Generative Synthesis Of Historical Precedents For Automated Procedural Guidance, Pooja Agrawal, Tanmay Kayande, Akash Dambhare, Deep Raithatha, Mohith P

Defensive Publications Series

Knowledge workers in enterprise environments can face challenges when resolving complex inquiries, as they may need to manually parse and synthesize information from multiple historical precedents. Existing retrieval systems may identify relevant cases but can leave the cognitive burden of synthesis to the user. Systems are described that can automate the synthesis of insights. For example, a system may use a semantic similarity model to retrieve a set of procedurally analogous historical cases from a data repository. A generative synthesis module could then compile data from these retrieved cases into a structured prompt for a large language model. The model …

Remote Procedure Call Performance Service For Performance Testing, Reda Nassif Haddad

Defensive Publications Series

A new remote procedure call (gRPC) performance service is proposed herein, referred to as a gRPC-Perf service, which can enable a network operator to understand gRPC limitations for any gRPC service that the network operator intends to use between any two points in a network. The gRPC-Perf service may also be useful for troubleshooting gRPC performance issues (e.g., whether an issue is a client issue or a network issue based on direction of traffic).

On-Chip Thermal Monitoring And Mitigation For Semiconductor Design-For-Test, Veerabhadra Rao Vasa, Maheedhar Jalasutram, Aditya Kota, Swathi G

Defensive Publications Series

During semiconductor design-for-test operations, the application of high-activity test patterns can generate localized heat as functional-mode thermal management systems may be inactive. A hardware-based architecture can integrate localized monitoring modules near potential thermal hotspots. A module can operate synchronously with test patterns to capture peak temperature values and their corresponding test pattern counts. The system may also include a mitigation circuit that compares real-time temperature against a configurable threshold and can throttle system clocks if the threshold is exceeded to reduce heat generation. This approach can provide thermal observability and control in a test environment, enabling the creation of thermal …

Automated Database Selection Via Empirical Validation And Cost-Adjusted Scoring, Nazar Ahamed, Navakanth Talluri

Defensive Publications Series

Selecting a database service can be challenging when decisions rely on static information or generalized benchmarks, which may not align with an application's specific runtime performance needs, potentially leading to implementation risks. A system can provide automated, performance-validated database recommendations. The system can translate a user's non-functional requirements into a profile used to programmatically provision temporary test environments for candidate databases. A workload generator can then simulate a user's specific application load within these environments, which may allow the system to harvest real-time performance and cost metrics. The collected empirical data can be processed by a scoring algorithm to calculate …