AI/ML arXiv cs.AI

A Tutorial on Autonomous Fault-Tolerant Control Using Knowledge-Grounded LLM Agents

A tutorial on using knowledge-grounded LLM agents for autonomous fault-tolerant control in process plants, including executable Python environments for testing.

AI/ML arXiv cs.AI

Intrinsic decomposition and editing of 3D Gaussian splats

A method for intrinsic decomposition and editing of 3D Gaussian splats, allowing for texture modification without altering scene lighting.

Cybersecurity arXiv cs.AI

A Lifecycle and Application-Stack Survey of Large Language Model Vulnerabilities: Attacks, Risks, Defenses, and Open Problems

A comprehensive survey of LLM vulnerabilities across the entire application stack, from data collection to deployment and agent execution.

Other Hacker News

Global review confirms mRNA vaccines are safe, effective and full of promise 

A global review confirms that mRNA vaccines are safe and effective.

Open Source Hacker News

LibreCAD in the Browser

LibreCAD, an open-source 2D CAD software, is now available to run in the browser.

AI/ML arXiv cs.AI

On the Convergence of Self-Improving Online LLM Alignment

Researchers introduce SAIL-RevKL, a regularized objective for the SAIL algorithm to ensure global convergence and improve LLM alignment stability.

AI/ML arXiv cs.AI

Improving multichannel speech enhancement through accurate room-acoustic simulations

Study demonstrates that using high-fidelity wave-based room-acoustic simulations significantly reduces word error rates in multichannel speech enhancement.

Cybersecurity arXiv cs.AI

CVE-TTP KG: Knowledge Graph Linking Software Vulnerabilities to Attack Behaviors

The CVE-TTP Knowledge Graph links software vulnerabilities (CVEs) to attacker behaviors using CySecBERT and Neo4j for better threat intelligence.

Open Source arXiv cs.AI

FLARE-AI: Flaw Reporting for AI

FLARE-AI is introduced as an open-source, interoperable system for reporting and triaging flaws in deployed AI systems to improve safety.

AI/ML arXiv cs.AI

Mitigating Positional Leakage in 3D Masked Autoencoders for Robust Representation Learning

MPL-MAE is a 3D Masked Autoencoder framework that mitigates positional leakage to improve semantic representation learning in point clouds.

AI/ML arXiv cs.AI

Temperature Field Reconstruction of Tungsten Monoblock Divertor on EAST using Physics-aware Neural Operator Transformer

Physics-aware Neural Operator Transformer (PNOT) is proposed for real-time temperature field reconstruction of fusion device divertors.

AI/ML arXiv cs.AI

DPPE: Rethinking Camera-Based Positional Encoding for Scaling Multi-View Transformers

Decoupled Pose Positional Encoding (DPPE) is proposed to resolve training bottlenecks in multi-view Transformers for novel view synthesis.

AI/ML arXiv cs.AI

ZEBRA: Zero-Shot Entropy-Regularized Prompt Learning for Base-to-Novel Generalization in Audio-Language Models

ZEBRA is a plug-and-play framework for audio-language models that reduces the base-to-novel generalization gap using self-entropy regularization.

Software Engineering Hacker News

The <Usermedia> HTML Element

A discussion about the <usermedia> HTML element, a legacy feature designed for user-uploaded content.

Software Engineering Hacker News

Show HN: Curvytron 2, I rewrote my browser party game, 10 years later

A developer shares a rewritten version of their browser-based party game, Curvytron 2, developed ten years after the original.

Tech Business/VC Hacker News

Meta Caps Internal AI Token Spending After Costs Approach Billions in 2026

Meta is reportedly capping internal AI token spending as projected costs are expected to reach billions by 2026.

AI/ML arXiv cs.AI

Temporal Preservation over Processing: Diagnosing and Designing Spatiotemporal Single-Stage Video Detectors

Researchers introduce TemporalLens for diagnosing temporal dependence in video detectors and propose YOLO-3D, a spatiotemporal detector based on YOLOv8.

AI/ML arXiv cs.AI

DA-Studio: An Agentic System for End-to-End Data Analysis

DA-Studio is an interactive, sandboxed web-based system designed for autonomous, end-to-end data analysis using LLMs.

AI/ML arXiv cs.AI

UniTac: A Unified Multimodal Model for Cross-Sensor Tactile Understanding and Generation

UniTac is a unified multimodal model for understanding and generating tactile signals across different sensors.

AI/ML arXiv cs.AI

Team MKC at CLPsych 2026: Capturing and Characterizing Mental Health Changes through Social Media Timeline Dynamics

A proposed LLM-based pipeline for analyzing mental health changes over social media timelines as part of the CLPsych 2026 shared task.