Digital Commons Network^™

Tunable Multi-Stage Query-Based Protocol For Personalized Content Recommendations, Animesh Sinha, Jital Patel, Akshay Gaur

Defensive Publications Series

Content recommendation systems rely on user-behavior statistics to model preferences. These systems must be engineered to handle an ever-growing corpus of documents, a vast user base, and sparse signals for inferring latent interests. Additionally, they face significant challenges in determining the appropriate dimensionality and granularity for interest modeling. To produce high-quality systems that function at web-scale, it is essential to leverage global semantic knowledge. Conventionally, this is achieved through various embedding and tagging systems that provide a foundational vocabulary for interest modeling. In this work, we propose a multi-stage, query-based architecture that leverages Large Language Models (LLMs) to learn high-quality …

Dynamic Configuration Of Artificial Intelligence Agent Contexts Via Shortened Aliases, Tomek Rutowski

Defensive Publications Series

The present disclosure relates to systems and methods for configuring the operational context of artificial intelligence (AI) agents using short, memorable text phrases. Large language models (LLMs) and agentic models often require precise contextual scoping to function effectively, as unrestricted access to vast data sources can lead to hallucinations or broad, inaccurate responses. The described technology utilizes a configuration alias, similar to a shortened uniform resource locator (URL), which is recognized by a parser within the AI interface. Instead of processing this alias as a standard prompt, the system performs a backend lookup to retrieve a pre-defined set of configuration …

Style-Preserved Inference Flow (Spif): Real-Time Style-Locked Image Generation Through Model Pruning And Latent Space Optimization, Douglass W. Sharp

Defensive Publications Series

This publication discloses Style-Preserved Inference Flow (SPIF), a system for real-time style-locked image generation that maintains consistent artistic style across frames at interactive rates. The core innovation is structural model pruning guided by style-consistency metrics rather than general image quality, producing compact single-purpose models that generate images in a specific artistic style at 60+ frames per second. Additional disclosed techniques include latent-space frame coherence (bootstrapping sequential frames from previous latent states, analogous to dirty-rectangle optimization in game rendering), composable latent object descriptors for real-time scene manipulation, and voice-driven continuous generation loops. Working prototypes demonstrate 3.35 FPS full-pipeline generation at 1024×1024 …

A Stateful, Hypothesis-Driven Framework For Long-Context Root Cause Discovery In Telemetry Analysis, Aakash Phoughat

Defensive Publications Series

This disclosure describes techniques that leverage a large language model (LLM) to perform root cause analysis (RCA) on high-volume telemetry data. The techniques can handle datasets whose size exceeds the context window of the LLM by traversing the data using a stateful log-walker that recursively passes an investigation state between LLM inference calls. A vector search is executed to retrieve a domain-specific structured graph of expected state transitions. The log stream is ingested in sequential blocks. A mutable investigation state which is a structured object that tracks the validation status of the steps of the schema is maintained. The object …

Screenshot Tokenization Guided By User Interface Tree, Florian Hartmann, Duc-Hieu Tran

Defensive Publications Series

Image tokenization is a technique that divides an image into multiple patches and embeds each patch into a vector space. Image tokenization is important for large language models (LLMs) to effectively answer queries relating to an image. A limitation of current image tokenization techniques for screenshots is that the patches are chosen in a manner that does not take into account user interface semantics, resulting in low information efficiency of tokens and user interface (UI) elements being split across tokens. This disclosure describes techniques that leverage UI element trees to guide screenshot tokenization, leading to higher quality screenshot tokens and …

Physical Design With Diagonal Cut Floorplan To Resolve Routing Congestion Hotspots, Sachin Ramrao Waghmare

Defensive Publications Series

In VLSI physical design, a Hard Macro (HM) instantiation with a vertical and horizontal cut can often intrude into the logic placement area. Typically, all routing layers are blocked in HMs. As a result, any nets passing through the area face a routing resource crunch that can cause routing congestion and design rule check (DRC) failures. This disclosure describes techniques for creating an HM instantiation during the PD phase of a digital design with a diagonal cut. The HM instantiation is cut using small notches formed by short vertical and horizontal cuts, which cumulatively provide a cut that is effectively …

Comprehensive Framework For Artificial Intelligence Model Testing: Integrated Validation Methodologies, Global Regulatory Compliance, And Enterprise Implementation Strategies For 25 Years, Ramkumar Bharathan

Defensive Publications Series

The proliferation of artificial intelligence systems across critical infrastructure necessitates a paradigm shift from traditional software quality assurance toward probabilistic validation methodologies. This paper presents a comprehensive framework for AI model testing that synthesizes regulatory requirements across multiple jurisdictions—including the EU AI Act (Regulation 2024/1689), NIST AI Risk Management Framework 2.0, Federal Reserve SR 11-7, Colorado AI Act, NYC Local Law 144, and sector-specific regulations—into an integrated validation architecture applicable across financial services, healthcare, insurance, manufacturing, and retail sectors.

The framework encompasses twelve distinct testing dimensions: functional correctness, performance benchmarking, bias and fairness assessment, robustness and adversarial resilience, explainability validation, …

Prismatic Subspace Projection For Semantic Control Of Vector Embeddings, Thomas Joseph Duerig, Yichang Chen, Jiří Iša

Defensive Publications Series

Vector search effectiveness may be reduced by semantic mismatches where nuisance dimensions, such as formality or complexity, can degrade retrieval relevance. Addressing these dimensions could involve resource-intensive model retraining. Systems and methods are described for dynamically modifying vector embeddings using prisms, which can be vector representations of certain semantic concepts. For example, a prism can be calculated from the vector difference between embeddings of two data items that differ primarily along a given semantic axis. To refine a search, a query or document embedding can be modified through a geometric projection into a subspace orthogonal to a prism vector. This …

Continuous Data Streaming Via A Jtag Interface, Harsharaj Ellur

Defensive Publications Series

Traditional JTAG data transfers often face limitations due to the “stop-and-wait” nature of the IEEE 1149.1 protocol. High overhead is introduced because the Test Access Port (TAP) state machine must be transitioned between SHIFT-DR and UPDATE-DR states for every individual data word. This disclosure describes a JTAG-to-AXI bridge microarchitecture that implements a streaming-write mechanism. A Command FIFO is utilized to decouple JTAG programming from AXI bus execution. An internal bit-shift counter and control logic are employed to automatically push data into a Write FIFO after a fixed number of bits are shifted. This allows a JTAG host to remain in …

Visual Movement Indicator For Head-Worn Displays, Chia Wei Kao

Defensive Publications Series

Collisions between users of head-worn displays and individuals in the immediate environment often occur because the physical movement of the user is unpredictable to bystanders. This disclosure describes a head-worn display equipped with external visual indicators and an integrated orientation sensor, such as a gyroscope. The sensor detects changes in head position or user leaning, and specific patterns of light are emitted from the visual indicators based on the detected direction. For example, forward, backward, or lateral movements trigger corresponding indicators on the front, rear, and/or sides of the device chassis. By providing real-time visual signals of a user's intended …

Quantitative Importance Attribution For Source Documents In Generative Models, N/A

Defensive Publications Series

Generative models that utilize external information sources may not include mechanisms to quantitatively measure the contribution of each source document to a generated output. Disclosed are systems and methods for quantitatively attributing source importance. A technique can involve calculating numerical importance scores for constituent parts of a model's output, for instance, by analyzing internal model attention scores or semantic similarity to a user prompt. A remapping process then can trace these scored parts back to their original source documents to assign an overall importance score to each source. This quantitative attribution data can provide a basis for feedback mechanisms for …

Just-In-Time Adaptive Interventions For Wellness Management Via Context-Aware Wearable Devices, Cecilia Abadie, Aveek Purohit, Pinal Bavishi, Theo Guidroz, Ayush Jain, Ines Mezerreg

Defensive Publications Series

ABSTRACT

Manual logging for health and wellness often results in low user engagement due to the high effort and friction involved for data entry. Additionally, ongoing activity observation using high-fidelity sensors on wearable devices is constrained by finite battery life and computational resources. To address these challenges, a unified architecture is described that facilitates just-in-time adaptive interventions.

A cascaded sensing layer is utilized to reduce power consumption. A low-power sensor screens for motion signatures indicative of potential events, such as hand-to-mouth gestures, on a persistent basis. Upon detection, a high-power sensor is triggered to capture rich contextual data. This data …

Segmented Touchpad Interface For Text Input On Smart Eyewear, Zhi Li, Xuelin Huang

Defensive Publications Series

Text entry on head-mounted wearable devices is often constrained by compact interfaces, leading to slow or inaccurate input. To address these limitations, a segmented touchpad interface is provided on the temple of the wearable device. A traditional keyboard layout is split into four distinct regions, which are mapped to four corresponding segments on the physical touchpad. Each segment is configured to represent multiple characters, and text is entered through a sequence of taps on these regions.

Linear swipes along the touchpad are utilized to replicate rotational input for selecting suggestions, while long-presses and glides enable character selection from pop-up menus. …

Process For The Preparation Of (4as,6r,8as)-4a,5,9,10,11,12-Hexahydro -3-Methoxy-11-Methyl-6h-Benzofuro[3a,3,2-Ef] [2]Benzazepin-6-Benzoate Gluconate, Msn Laboratories Private Limited, R&D Center; Srinivasan Thirumalai Rajan, Revu Satyanarayana, Dr. Nakka Mangarao, Sunkari Suresh Kumar, Gunde Mahender.

Defensive Publications Series

The present invention relates to a process for the preparation of (4aS,6R,8aS)-4a,5,9,10,11,12-hexahydro-3-methoxy-11-methyl-6H-benzofuro[3a,3,2-ef][2]benzazepin-6-benzoate gluconate.

Retrieving Personalized Content Using Ai Agents And User Profile Information, Benjamin Azose, Akancha Gupta

Defensive Publications Series

Agentic workflows can automate many user journeys. However, when a user journey involves choosing between different options, e.g., dietary preferences, the AI agent may be inadequate in automating the user journey, requiring manual actions and/or user inputs to make a choice. This disclosure describes techniques that automate agentic workflows by permitting an AI agent to access user preferences from a user profile prior to performing a corresponding task. With user permission, user preferences are obtained via explicit input, user browsing history, transactions, etc. and are stored in a user profile that is accessible to AI agents. When summoned to perform …

Measuring Entity Perception From User-Generated Video Content Using Multimodal Language Models, Angel Raposo, Prakhar Rathi

Defensive Publications Series

A system related to measuring entity perception from user-generated video content using multimodal large language models. The system implements a multi-stage data ingestion pipeline using Knowledge Graph ID filtering to identify relevant videos, processes videos through a multimodal language model to generate attribute-specific perception scores with rationales, aggregates scores across temporal intervals with coverage metrics, and generates comparative visualizations with automated statistical analysis.

Semantic Navigation Using Generative Models And Graphical Maps, Mbhb Mbhb, Artemis Panagopoulou, Aveek Purohit, Achin Kulshrestha, Soroosh Yazdani, Mohit Goyal

Defensive Publications Series

Indoor navigation presents a significant challenge because Global Positioning System (GPS) accessibility is limited in interior spaces and static environmental scans often fail to generalize as environments change. To address these limitations, a novel methodology utilizes generative models to produce diverse map layouts and vision-language models (VLMs) to interpret potentially walkable areas within those layouts. By scaling this approach to generate a large dataset, significant improvements are achieved in the wayfinding capabilities of the models. This technology is particularly applicable to navigational agents and wearable devices, such as glasses, that require real-time semantic understanding of human-readable graphical maps. The resulting …

Advanced User Interaction Logging For Recommendation Model Enhancements, Tamojit Chatterjee, Kanishk Mishra

Defensive Publications Series

Traditional user interaction logging processes typically record discrete events based on user interactions with a user interface of an application. The user interactions may include user clicks or selections of interactive user interface elements (e.g., click events) and/or user impressions of content (e.g., impression events) provided by the user interface. Interactive user interface elements may include, but are not limited to, buttons, selectable icons, advertisement cards, and content cards. For example, a user impression may occur when a content item (e.g., an advertisement, an image, a video, etc.) is rendered in the user interface (e.g., in a respective content card) …

A Method For Calibrating Head-Worn Systems, Mar Gonzalez Franco, Karan Ahuja, Qiao Yang, Eric Jordan Gonzalez, Andrea Colaco, Khushman Jayantilal Patel, Prasanthi Gurumurthy

Defensive Publications Series

This disclosure describes a solution for calibrating coordinate frames of a head-worn device. Unlike existing eye-tracking systems in head-worn systems, such as mixed reality (MR)/augmented reality (AR) glasses or head-mounted displays (HMDs), that require disruptive calibration, such as following virtual dots, this approach correlates a user’s gaze with real-world objects. By utilizing scene understanding (e.g., simultaneous localization and mapping (SLAM) or object recognition) to map the physical environment, the device prompts the user to gaze at a real-world object or audio source. The system then monitors the user's gaze and determines a spatial offset between the tracked gaze and the …

Gradient-Based Explainability For Functional Neural Networks Via Differentiable Layers, Aniruddha Rao, Chetan Sharma, Chenyin Gao, Fernando Rodriguez Silva Santisteban

Defensive Publications Series

A system and method are described related to the interpretability of predictive models, such as functional neural networks (FNNs), that can operate on continuous data. For some FNNs, certain explainability techniques may be less effective as they can involve discretizing an input signal, which may result in a loss of functional information. The disclosed technology can integrate custom, differentiable functional layers, such as basis expansion or inner product layers, within the neural network architecture. This design can preserve a differentiable path from a model's output to the original functional input, which may enable the application of gradient-based attribution methods. This …

A Process For The Purification Of Abemaciclib, Anonymous

Defensive Publications Series

The present publication relates to a process for the purification of Abemaciclib, chemically known as N-[5-[(4-ethyl-l-piperazinyl) methyl]-2-pyridinyl]-5-fluoro-4-[4-fluoro-2-methyl-1- (1-methylethyl)-1H-benzimidazol-6-yl]-2-pyrimidinamineof Formula-1.

The Gossett Protocol: A 120-Point Master Engineering Standard For Sovereign Neural-Interface Bypasses, Jon S. Gossett

Defensive Publications Series

This disclosure establishes a comprehensive 120-point industrial engineering standard for a semi-invasive digital spinal bypass and neural communication bridge. The protocol prioritizes "Human Sovereignty" through a hardware-locked, 16-bit discrete metadata standard, ensuring all motor intent is processed locally without cloud dependency. Key technical innovations include a 1.0mm non-penetrative cortical clearance zone, a medical-grade 850nm NIR optical bridge for transdermal signal transmission, and a 38.5°C thermal breach auto-kill safety wall. Version 6.2 introduces Category IX, detailing a "Camera-Assisted Grounding" process where external optical sensors are utilized exclusively for initial dictionary calibration and intent verification. This optical requirement is phased out once …

---

Go to article

Responsive User Interfaces Based On Task Criticality And User Context, Piyush Arora, Colby Hawker, Shantanu Pai, Ram Vivekananda, Wendy Yun, Jaime Sonoda

Defensive Publications Series

Responsive user interfaces enable dynamically adjusting user interfaces based on device-specific aspects such as screen size, aspect ratio, display resolution, etc. However, traditional responsive design fails to account for different types of constraints of a user and task criticality of the task being performed via the UI. Misalignment between the UI design, user context and task criticality can lead to user error. This disclosure describes techniques, implemented with user permission, for dynamically modifying the layout, information density, and/or interactive physics of a user interface based on a dual-factor analysis of user cognitive state and task criticality. The user's cognitive state …

Methods And Systems For Modifying Intron Sequences To Remove Restriction Sites While Preserving Gene Accumulation Enhancement In Transgenic Plants, Cory Tobin

Defensive Publications Series

The present invention provides methods and systems for modifying intron sequences to enable their use in modular cloning (MoClo) systems while preserving their ability to enhance gene accumulation. The invention addresses a critical challenge in molecular cloning where beneficial introns contain restriction enzyme recognition sites that interfere with MoClo assembly. The methods include computational approaches for identifying minimal sequence modifications needed to eliminate restriction sites while maintaining intron functionality, experimental validation protocols for confirming preserved enhancement properties, and high-throughput screening systems for rapidly assessing multiple intron variants. In particular embodiments, the invention demonstrates successful modification of introns for enhanced accumulation …

Ground-Isolation Power Switching For Multi-Source Dc Inputs Via Mechanical Shunt, Pellumb Bara

Defensive Publications Series

This disclosure describes a simplified power-management circuit for electronic devices with dual power inputs (e.g., USB-C and DC Barrel Jack). It utilizes the internal mechanical shunt of a standard DC barrel jack to physically disconnect the ground path of the secondary power source (USB-C) when an external power adapter is inserted. This provides galvanic isolation between power sources without the need for active semiconductors or complex comparator circuits.

The Online Advertising System That Provides Web Services And Web Related Services Where The Advertising Revenue Is Utilized In The Generation Of The Providing Services., Punarjeewa Abeysekera

Defensive Publications Series

This paper describes an online advertising system that provides web services and web related services where the advertising revenue is utilized in the generation of the providing services. The online advertising system consists of a server facility that have the capability to receive and provide content online, a group of advertising content providing business entities and a group of advertising content watching and free web services and free web related services consuming client entities. An entity who will utilize the online advertising system can concurrently function as an advertising content providing business entity and as an advertising content watching and …

Belief Hijacking Attacks On Ai Defenders: “Manipulating What Ai Systems Believe About Threats, Actors, And Risks”, Pranav Bhatanagar Mr

Defensive Publications Series

Artificial intelligence systems are increasingly deployed as autonomous defenders in modern cybersecurity environments. These systems continuously analyze network behavior, evaluate threat intelligence, assess risk levels, and recommend or execute defensive actions. Central to their effectiveness is the formation of internal beliefs regarding attacker intent, infrastructure trustworthiness, vulnerability severity, and operational priority. While existing research has focused on attacks that manipulate detection accuracy, evade classifiers, or bypass response mechanisms, far less attention has been given to vulnerabilities in belief formation processes themselves. This paper introduces Belief Hijacking Attacks, a novel class of adversarial strategies that target how AI defenders construct, update, …

Trust Amplification Exploits In Human To Ai Security Teams Abusing Overconfidence In Automated Recommendations, Pranav Bhatanagar Mr

Defensive Publications Series

Human–AI collaboration has become a central component of modern cybersecurity operations. Security analysts increasingly rely on automated systems for threat detection, risk assessment, and incident response. While this collaboration improves efficiency and scalability, it also introduces new vulnerabilities associated with human trust in machine generated recommendations. This paper introduces Trust Amplification Exploits, a novel class of adversarial strategies that manipulate and exploit overconfidence in AI-assisted security workflows. Rather than targeting algorithms directly, these attacks leverage repeated system accuracy, interface design, and organizational dependence to amplify human reliance on automated outputs. Over time, excessive trust reduces critical evaluation and increases susceptibility …

Reality Distortion Attacks On Autonomous Security Agents: Manipulating Internal World Models In Ai Defenders, Pranav Bhatanagar Mr

Defensive Publications Series

Autonomous artificial intelligence systems are increasingly deployed as primary defenders in modern cybersecurity environments. These systems rely on internal world models to interpret network behavior, assess threats, and guide automated response. While existing research has focused extensively on attacks that manipulate inputs, outputs, and learning processes, limited attention has been given to threats that target internal perception. This paper introduces Reality Distortion Attacks, a novel class of adversarial strategies that manipulate how autonomous security agents model and understand their operational environment. Rather than inducing immediate misclassification, these attacks gradually reshape situational awareness by influencing sensor inputs, feedback mechanisms, contextual signals, …

Epistemic Corruption Attacks: Poisoning What Ai Security Systems “Know” Rather Than What They “Do”, Pranav Bhatanagar Mr

Defensive Publications Series

As artificial intelligence becomes central to modern cybersecurity operations, defensive systems increasingly rely on internal knowledge representations, long-term memory modules, and retrieval-based reasoning mechanisms. These components enable adaptive learning and contextual awareness, but they also introduce a largely unexamined security vulnerability: the integrity of what the system knows. Existing research on AI security has primarily focused on attacks that manipulate outputs, bypass detection mechanisms, or compromise execution pathways. Far less attention has been given to attacks that corrupt internal knowledge structures themselves. This paper introduces Epistemic Corruption Attacks, a novel class of adversarial strategies that target the belief systems, embeddings, …