AI/ML arXiv cs.AI

Music-to-Dance Generation via Atomic Movements

Presents a structure-aware framework for music-to-dance generation that uses atomic movements and LLMs for semantically interpretable choreography.

Tech Business/VC arXiv cs.AI

The Dynamic Verifiable Multi-Agent Human Agentic Loyalty Loop (DVM-HALL) Model and the Net Human-Agent Score (NHAS) in Autonomous Commerce

Proposes the DVM-HALL model and NHAS metric to evaluate brand loyalty and trust in autonomous commerce where AI agents make purchasing decisions.

Cybersecurity arXiv cs.AI

Rethinking Penetration Testing for AI-Enabled Systems: From Resource Compromise to Behavioral Objective Violation

Reframes penetration testing for AI systems to focus on behavioral objective violations rather than just infrastructure compromise, covering prompt injection and data poisoning.

AI/ML arXiv cs.AI

Transforming Rank: How Architecture Navigates the Spectral Pathologies of Depth

Analyzes how Transformer architecture components like skip connections and normalization preserve or collapse gradient rank across network depth.

AI/ML arXiv cs.AI

Improving Wind and Solar Power Prediction with Efficient Wrapper-based Feature Selection: An Empirical Study

Proposes Cluster-based Sequential Feature Selection (CSFS) to improve wind and solar power prediction efficiency and provides an open-source implementation.

AI/ML arXiv cs.AI

Early Adoption of Agentic Coding Tools by GitHub Projects

An empirical study of agentic coding tools on GitHub, finding that adoption is concentrated in small projects and typically follows a single-human oversight model.

AI/ML arXiv cs.AI

Anatomically Faithful but Temporally Blind: Auditing Attribution for Left-Ventricular Ejection-Fraction Estimation from Echocardiography

Researchers found that AI models for estimating heart ejection fraction are anatomically accurate but fail to identify the clinically decisive frames in a video, suggesting a need for temporally-aware training.

AI/ML arXiv cs.AI

MxGPS: Multiplex Graph Transformers for a Power Grid Foundation Model

MxGPS is a multiplex graph transformer designed for power grid models that prevents 'topology overfitting,' enabling better generalization to unseen grid structures.

Hardware/Chips arXiv cs.AI

Kaleido: Algorithm-Hardware Co-Design for Video Diffusion Transformers by Exploiting Latent Space Correlations

Kaleido is an algorithm-hardware co-design that accelerates Video Diffusion Transformers by exploiting spatiotemporal correlations, achieving up to 5.9x speedup.

AI/ML arXiv cs.AI

CAS I: A Geometric Coding Theorem

This paper presents a Geometric Coding Theorem that unifies algorithmic information theory with group theory to study symmetry-induced complexity measures.

Cybersecurity arXiv cs.AI

Traffic-Aware Randomized Smoothing for LLM-Based Network Intrusion Detection

The authors propose Traffic-Aware Randomized Smoothing (TA-RS) to provide certified robustness for LLM-based network intrusion detection systems against traffic manipulation.

AI/ML arXiv cs.AI

Multimodal Assessment of Pancreatic Cancer Resectability Using Deep Learning

A new multimodal deep learning framework using Swin-UNETR jointly analyzes CT images and clinical data to automate pancreatic cancer resectability classification.

AI/ML arXiv cs.AI

NodeImport: Imbalanced Node Classification with Node Importance Assessment

NodeImport is a framework for imbalanced node classification in graphs that uses a balanced meta-set to identify and prioritize significant nodes for training.

Tech Business/VC arXiv cs.AI

AI-Augmented Human Resource Management? Insights from German companies

A study of German companies reveals that while AI is being integrated into HR for efficiency, it often focuses on rationalization rather than strategic talent development.

AI/ML arXiv cs.AI

Unleashing Multimodal Large Language Models for Training-free HOI Detection in the Wild

AgentHOI is a training-free agentic framework that uses multimodal reasoning and spatial grounding for open-world human-object interaction detection.

AI/ML arXiv cs.AI

Verifying formulas for interventional distributions

The authors formalize the verification of observational formulas in causal graphical models to determine if they correctly identify target interventional distributions.

Software Engineering Hacker News

Solod: Go can be a better C

A discussion on the potential for the Go programming language to serve as a more effective alternative to C for systems programming.

Software Engineering Hacker News

Scaling to 1M concurrent sandboxes in seconds

An exploration of techniques to scale sandboxes to handle one million concurrent instances within seconds.

Other Ars Technica

Two Trump health nominees crash and burn in tense Senate hearing

Reports on the poor performance of two Trump health nominees during a tense Senate hearing.

Tech Business/VC Ars Technica

HP fined 1.4 billion rupees for “cartelization” of ink cartridges, toner, PCs

HP was fined 1.4 billion rupees for engaging in cartelization practices related to ink cartridges, toner, and PCs.