All Articles
17198 articles total
Automatic Thematic Indexing of Large Literary Corpora: A Machine Learning Approach to Voltaire's Complete Works
Research on using machine learning, specifically 4-bit quantized Mistral models with LoRA, for the automatic thematic indexing of Voltaire's literary works.
ReGen: Hierarchical Multi-Prompt Representation Generation for Efficient Waveform Diffusion Models
ReGen is a hierarchical multi-prompt representation generation framework that improves waveform generation quality in diffusion models, specifically for text-to-speech.
A Personalized Computational Framework for Assessing the Sufficiency of Partially Observed Data in Healthcare AI models
The authors introduce Feature Sufficiency Analysis (FSA) to determine if a subset of available clinical features is enough for an AI model to reach full-feature-capacity in healthcare.
Attention to Detail: Evaluating Energy, Performance, and Accuracy Trade-offs Across vLLM Configurations
A large-scale study on vLLM configurations reveals that attention kernel type and prefix caching significantly impact energy, performance, and occasionally accuracy.
Generative Communications: Overview, Technologies, and Trends
This paper proposes Generative Communications (GenCom), a 6G network paradigm where large AI models drive semantic understanding and content generation rather than simple bit transmission.
Interference and Retention in Continual Learning
The paper introduces Interference-Gated Functional Allocation (IGFA), a replay-free method to eliminate forgetting in continual learning by modeling forgetting as task interference.
Tactile and Vision Conditioned Contact-Centric Control for Whole-Arm Manipulation
TACTIC is a receding-horizon controller for whole-arm robot manipulation that combines RGB-D, tactile sensing, and proximity representations for better contact-centric control.
Git-Assistant: Planning-Based Support for Updating Git Repositories
Git-Assistant combines LLMs with automated planning to help developers execute complex git operations more reliably than using LLMs alone.
All you need is SAMPAT
SAMPAT is a three-layer neural architecture that provides fully interpretable, closed-form algebraic expressions while remaining competitive with deep neural networks.
LLMs for health: Perceived benefits, risks, intention to use AI chatbots, and willingness to self-disclose across sensitive health topics
A study on Dutch participants' perceptions of AI chatbots for health reveals that usage intention is driven by perceived benefits and risks rather than the topic sensitivity.
Blockchain-Linked Auditable Decision Management for Telecom/IoT Fraud-Control Requests
The paper proposes a blockchain-linked framework for auditing telecom/IoT fraud-control decisions using a combination of QLoRA-tuned LLMs and centralized ML.
Tracking unique visitors without cookies
A Hacker News discussion regarding methods for tracking unique visitors to a website without relying on browser cookies.
Inside the Skill Market: From Software Engineering Activities to Reusable Agent Skills
A large-scale empirical study analyzing how software engineering activities are being encapsulated into reusable 'skills' for AI agents.
OmniMapBench: Benchmarking Visual-Centric Reasoning on Diverse Map Documents
Introduction of OmniMapBench, a benchmark for testing visually-centric reasoning in Large Vision-Language Models (LVLMs) using map documents.
PRecG: Legal Precedent Retrieval with Graph Neural Networks and Rhetorical Role Segmentation
PRecG is a pipeline for legal precedent retrieval that uses Graph Neural Networks and rhetorical role segmentation to improve semantic similarity search.
A Coreset Selection Framework with Ensemble Aggregation for Image Classification
A coreset selection framework called SCOSS that uses ensemble aggregation to improve image classification efficiency and training costs.
Beyond Metadata: CAPRA for Hidden Subgroup Analysis under Missing Metadata in Medical Imaging
CAPRA, a framework for hidden subgroup analysis in medical imaging to identify failure modes when demographic metadata is missing.
Integrating Large Language Models and Graph Convolutional Networks for Semi-Supervised Image Classification
A method for semi-supervised image classification that integrates LLMs and Vision Language Models to refine Graph Convolutional Networks (GCNs).
Event Stream based Multi-Modal Video Anomaly Detection: A Benchmark Dataset and Algorithms
The EVAD framework and a large-scale benchmark dataset for multi-modal video anomaly detection using event cameras and visible light video.
Augmenting Fundamental Analysis with Large Language Models: A RAG-Based System for Generating Investor Briefs
A study on using LLMs and RAG to automate the generation of investor briefs based on SEC filings and macroeconomic data.