AI/ML arXiv cs.AI

AgentX: Towards Agent-Driven Self-Iteration of Industrial Recommender Systems

AgentX is a multi-agent system designed to automate and improve industrial recommender systems through self-iterating development loops.

AI/ML arXiv cs.AI

TAVR-VLM: Risk-Conditioned Causal Grounding for Hallucination-Resistant Report Generation

TAVR-VLM is a framework for hallucination-resistant medical report generation in TAVR planning using risk-conditioned causal grounding.

Open Source Hacker News

We All Depend on Open Source. We Will Defend It Together

A community discussion on Hacker News regarding the collective defense and support of open-source software.

AI/ML arXiv cs.AI

Do Safety Guardrails Need to Reason? LeanGuard: A Fast and Light Approach for Robust Moderation

Introduction of LeanGuard, a lightweight bidirectional encoder for robust AI moderation that achieves 100x reduction in inference compute compared to reasoning-based guardrails.

AI/ML arXiv cs.AI

Kalman Prototypical Networks for Few-shot Fault Detection in Combined Cycle Gas Turbines

Proposes the Kalman Prototypical Network (KPN) for few-shot fault detection in gas turbines, using latent stochastic states to improve robustness.

AI/ML arXiv cs.AI

LithoDreamer: A Physics-Informed World Model for Multi-Stage Computational Lithography

Presents LithoDreamer, a physics-informed world model for computational lithography to optimize semiconductor manufacturing processes.

AI/ML arXiv cs.AI

A Latent ODE Approach to Spatiotemporal Modeling of Cine Cardiac MRI

A latent ODE approach to spatiotemporal modeling of cardiac MRI to improve heart failure prediction beyond conventional markers.

AI/ML arXiv cs.AI

Socratic agents for autonomous scientific discovery in high-dimensional physical systems

Introduces AHOIS, a multi-agent AI scientist that uses Socratic questioning to autonomously discover physical explanations in high-dimensional systems.

AI/ML arXiv cs.AI

Scientific discovery as meta-optimization: a combinatorial optimization case study

Proposes a meta-optimization framework for scientific discovery that uses consensus objective aggregation to speed up 3-SAT algorithm discovery by 67x.

AI/ML arXiv cs.AI

EGG: An Expert-Guided Agent Framework for Kernel Generation

Presents EGG, an expert-guided agent framework for GPU kernel generation that achieves a 2.13x speedup over PyTorch through staged optimization.

AI/ML arXiv cs.AI

ResilPhase: Plug-and-Play Phase Mapping and Noise-Resilient Macro-Trajectory Extrapolation for Diffusion Acceleration

Introduces ResilPhase, a noise-resilient acceleration framework for Diffusion Transformers (DiTs) using barycentric Lagrange extrapolation in ODE space.

AI/ML arXiv cs.AI

Memory Depth, Not Memory Access: Selective Parametric Consolidation for Long-Running Language Agents

Study on memory depth for long-running AI agents, introducing EVAF, a selective parametric consolidation mechanism for goal persistence.

Other Hacker News

Falcon GX the most powerful brand engineering tool

A discussion thread regarding a brand engineering tool called Falcon GX.

Other Hacker News

Why are we so obsessed with lawns?

A general discussion thread questioning the societal obsession with lawns.

AI/ML arXiv cs.AI

Explainable Ensemble-Based Machine Learning Models for Detecting the Presence of Cirrhosis in Hepatitis C Patients

Research on using ensemble-based machine learning models, specifically Extra Trees, to detect cirrhosis in Hepatitis C patients with high accuracy.

AI/ML arXiv cs.AI

EvoOptiGraph: Weakness-Driven Coevolution via Graph-Based Structural Generation for Optimization Modeling

Introduction of EvoOptiGraph, a framework where data and models co-evolve to improve LLM performance in automating optimization modeling from natural language.

AI/ML arXiv cs.AI

A Multi-Level Validation and Traceability Framework for AI-Generated Telescope Scheduling Decisions

A multi-level validation and traceability framework designed to ensure the reliability and executability of AI-generated telescope scheduling decisions.

AI/ML arXiv cs.AI

Content-Based Smart E-Mail Dispatcher Using Large Language Models

An LLM-based agent system developed to automate the dispatching of emails to relevant WhatsApp groups in an engineering college setting.

AI/ML arXiv cs.AI

LLM-based Models for Detecting Emerging Topics in Service Feedback

A methodology combining quantized LLMs and human-in-the-loop oversight to detect emerging service quality issues in public sector customer feedback.

AI/ML arXiv cs.AI

Autoformalization of Agent Instructions into Policy-as-Code

An autoformalization pipeline that uses an LLM generator-critic loop to translate agent instructions into formally verified Cedar Policy Language policies.