AI/ML arXiv cs.AI

Explainable-by-Design Audio Deepfake Detection via Wiener-Hopf Linear Prediction

An explainable audio deepfake detection framework using Wiener-Hopf linear prediction and a lightweight CNN.

Software Engineering arXiv cs.AI

Multi-Perspective Agentic Program Repair via Code Property Graphs and Temporal Execution Graphs

CT-Repair is an agentic program repair framework using Code Property Graphs and Temporal Execution Graphs to fix Java bugs.

AI/ML arXiv cs.AI

Can Induced Emotion Bias LLM Behaviors in Sequential Decision Making?

Study on how induced emotion affects LLM sequential decision-making using the Iowa Gambling Task.

Other Hacker News

New era for Gibraltar with removal of border controls with Spain

Discussion regarding the removal of border controls between Gibraltar and Spain.

Software Engineering Hacker News

Neverclick: Desktop application for performing mouse actions with your keyboard

Neverclick is a desktop application that allows users to perform mouse actions using their keyboard.

Tech Business/VC TechCrunch

A SpaceX vet raised $65M to pull wire harnesses out of the Cold War era

A SpaceX veteran has raised $65M to modernize wire harness manufacturing for rockets and satellites.

Hardware/Chips The Verge

Samsung’s new foldable display is harder to crease and damage

Samsung introduces Flex Titanium display technology for foldables to reduce creasing and improve durability.

AI/ML arXiv cs.AI

IQA-T1: Tool-based Visual Evidence Reasoning for Image Quality Assessment

IQA-T1 is a tool-based reasoning framework for Image Quality Assessment that uses specialized analysis tools to provide interpretable visual evidence.

AI/ML arXiv cs.AI

Demonstration of the common dual-channel feature decoupling characteristic of front-door mediation causal inference methods in whole-slice image classification

Research proposing a dual-channel feature decoupling characteristic to improve causal inference in whole-slice image classification for digital pathology.

AI/ML arXiv cs.AI

ARDepth: Auto-regressive Monocular Depth Estimation with Progressive Visual Conditioning

ARDepth is a new auto-regressive monocular depth estimation model that progressively constructs depth representations across spatial scales.

AI/ML arXiv cs.AI

The Computational Basis of Confidence in Large Language Models

A study exploring the computational basis of confidence in LLMs, finding that answer logits often behave as readouts of a latent decision variable.

AI/ML arXiv cs.AI

An Omnilingual-ASR-Based Speech-LLM System for the 2nd MLC-SLM Challenge

An omnilingual ASR-based speech LLM system using a cascaded diarization-then-recognition approach for the 2nd MLC-SLM Challenge.

AI/ML arXiv cs.AI

Agent-Safety Evaluations as Load-Bearing Evidence: A Vendor-Neutral, Cross-Harness Reconstructability Metric

Proposal of a vendor-neutral reconstructability metric to evaluate the validity and evidence sufficiency of agent-safety evaluations.

Tech Business/VC The Verge

Spotify’s Daniel Ek is bringing his body-scanning clinics to the US

Spotify founder Daniel Ek is launching Neko Health, a body-scanning clinic startup, in the US, starting with New York.

Hardware/Chips The Verge

The PS6 sure sounds like a handheld

Speculation regarding the PS6 suggests it may be a handheld device, following hints from a Sony investor meeting and the phase-out of physical discs.

AI/ML arXiv cs.AI

Track, Rank, Crack: Epistemic Working Memory Scales Multi-Hop Reasoning in Language Agents

Researchers introduce SLEUTH, a framework using structured epistemic working memory to improve multi-hop reasoning in language agents.

AI/ML arXiv cs.AI

Code-MUE: Measuring Code LLMs' Uncertainty through Execution-based Semantic Interaction Graphs

Code-MUE is a black-box framework that measures Code LLM uncertainty by analyzing execution-based Semantic Interaction Graphs.

AI/ML arXiv cs.AI

The Sound of Absence: Audio-Language Embedding Models Struggle with Negation

Study reveals that audio-language embedding models struggle with negation, proposing the NegEval-Audio framework to address this gap.

AI/ML arXiv cs.AI

A Longitudinal Analysis of Public Discourse on AI Ethics in Education Using Twitter Data

A five-year longitudinal analysis of Twitter data shows pragmatic public acceptance of AI in education, though concerns about academic integrity persist.

AI/ML arXiv cs.AI

A Comparative Analysis of Institutional and Course Generative AI Policies within Higher Education: Implications for Instruction in Computing Education

Comparison of AI policies in higher education shows that institutional guidelines are generally more permissive than course-level implementation.