All Articles
17524 articles total
DiPhon: Diffusion on Graphons for Scalable Graph Generation
DiPhon is introduced as a diffusion framework on graphons for size-scalable graph generation, allowing models to generate larger graphs without retraining.
Gimitest: A Comprehensive Tool for Testing Reinforcement Learning Policies
Gimitest is an open-source framework designed to test the reliability and safety of single- and multi-agent reinforcement learning policies across various gym environments.
AnchorPrune: Relevance-Anchored Contextual Expansion for Visual Token Pruning
AnchorPrune is a training-free framework for visual token pruning in vision-language models, improving inference efficiency by preserving query-critical evidence and complementary context.
Intrinsic Green's Learning: Supervised Learning on Manifolds via Inverse PDE
Intrinsic Green's Learning (IGL) models target functions on manifolds as solutions to linear PDEs, utilizing an encoder to discover low-dimensional coordinate charts for efficient computation.
On the Principles of Deep Feedforward ReLU Networks
This research provides a theoretical analysis of deep feedforward ReLU networks, explaining how they use piecewise linear manifolds to divide input space and the mechanism of back-propagation.
Riemannian Geometry for Pre-trained Language Model Embeddings
The authors propose Riemannian Mean Pooling (RMP) to analyze the geometric structure of pre-trained language model embeddings for better interpretability and safety.
Making Implicit Preservation Intent Explicit in Conversational Image Editing
The paper introduces OCCUR-Bench for evaluating temporal preservation in conversational image editing and ReSpec, a framework to improve restoration fidelity using historical references.
Progressive Crystallization: Turning Agent Exploration into Deterministic, Lower-Cost Workflows in Production
Progressive Crystallization is a method to reduce AI agent costs by converting repeatedly validated agent behaviors into deterministic, low-cost workflows in production AIOps.
Complexity-Budgeted, Interaction-Aware Interpretable Model for Tabular Data
IAIML is an interpretable machine learning framework for tabular data that captures pairwise feature interactions while maintaining a small explanation budget.
Multiplication Beyond Groups: Stratified Fourier Mechanisms in Transformer Circuits
This study investigates how transformers learn non-invertible modular integer multiplication by partitioning input space into local algebraic regions using Fourier mechanisms.
Navigating Hierarchy: Hyperbolic Learning on Brain Graphs for Disorder Diagnosis
Hyperbolic Learning on Brain Graphs (HLBG) uses Lorentzian hyperbolic space and a Graph-aware Mamba model to diagnose brain disorders by modeling hierarchical network relationships.
In-browser programmable robot simulator
A programmable robot simulator that runs directly in the web browser, allowing for accessible robotics experimentation.
Why developers are ditching GitHub for Codeberg and self-hosting alternatives
A discussion on why developers are migrating from GitHub to decentralized or self-hosted alternatives like Codeberg.
Self-Supervised Pretraining Improves Cross-Site and Cross-Scale Robustness of Point Cloud Leaf-Wood Segmentation
Research on using self-supervised pretraining with Point-M2AE to improve the robustness of leaf-wood segmentation in forest point clouds.
Large Language Models (LLMs) and Generative AI in Cybersecurity and Privacy: A Survey of Dual-Use Risks, AI-Generated Malware, Explainability, and Defensive Strategies
A comprehensive survey on the dual-use risks of LLMs in cybersecurity, covering AI-generated malware and defensive strategies.
End-to-End LLM Flight Planning with RAG-based Memory and Multi-modal Coach Agent
Introduction of FRAMe, an LLM-based flight planning tool for eVTOL aircraft using RAG-based memory and a multi-modal coach agent.
Hybrid Least Squares/Gradient Descent Methods for MIONets
A new hybrid least squares/gradient descent (LSGD) method to accelerate the training of MIONets.
WAM-TTT: Steering World-Action Models by Watching Human Play at Test Time
WAM-TTT is a test-time training framework that steers robot foundation models using human videos via self-supervised video prediction.
Physics-guided spatiotemporal neural models for fuel density prediction
A physics-guided machine learning framework for fuel density prediction to improve fire forecasting using ConvLSTM and ViViT.
Multimodal Spatiotemporal-Frequency Fusion with Peak Enhancement for Cellular Traffic Forecasting
MSPF-Net, a multimodal framework for cellular traffic forecasting that integrates spatial, spectral, and urban news data.