Software Engineering Hacker News

Shipping Godot VR and Porting to PSVR2: A Partial Post Mortem

A post-mortem on the technical challenges and experiences of shipping Godot VR and porting it to PSVR2.

AI/ML arXiv cs.AI

Beyond Predictive Accuracy: A Reliability-Aware Audit of Molecular Representations for Human Olfaction

A study auditing the reliability of molecular representations for human olfaction, finding that learned embeddings often fail to outperform conventional baselines.

AI/ML arXiv cs.AI

DisasterTD: Disaster Toponym Disambiguation Using Multimodal LLMs and Cross-View Geolocalization

Introduction of DisasterTD, a framework using multimodal LLMs and cross-view geolocalization for disambiguating disaster-related toponyms.

AI/ML arXiv cs.AI

HVM-GraphRAG: Holistic-View Multimodal Graph Retrieval-Augmented Generation on Complex Document

HVM-GraphRAG proposes a holistic-view multimodal GraphRAG framework to improve efficiency and accuracy in QA over complex documents.

AI/ML arXiv cs.AI

Tokens are All You Need: Dual-purpose Semantic IDs for Achieving LLM-Level I/O Efficiency in recommendation systems

A proposal for Dual-purpose Semantic IDs using hierarchical quantization to reduce the memory footprint and I/O overhead of large-scale recommendation systems.

AI/ML arXiv cs.AI

GraphRareBench: An Auditable Graph-Evidence Benchmark for Phenotype-Driven Rare-Disease Diagnosis

GraphRareBench is introduced as an auditable benchmark for phenotype-driven rare-disease diagnosis using graph evidence.

AI/ML arXiv cs.AI

Human Preference aligned Tabular Similarity

A paper arguing for the necessity of human-preference aligned evaluation for tabular embeddings used in similarity search.

AI/ML Hacker News

After the AI Crash

A community discussion on Hacker News regarding the potential 'AI crash' and the sustainability of current AI investments.

AI/ML Hacker News

Show HN: Bullshit Detector – agent skills that fact-check videos and articles

A new set of agent skills designed to fact-check videos and articles to detect 'bullshit' or misinformation.

AI/ML arXiv cs.AI

Retrieval-Augmented Generation in LLMs for Mental Health: Quantifying the Incremental Contribution of Retrieval Within a Layered Safety Architecture

Research exploring how Retrieval-Augmented Generation (RAG) improves intent detection and safety in LLMs used for digital mental health interventions.

AI/ML arXiv cs.AI

Dual-Level Atomic and Coordination Geometry Learning for Crystal Property Prediction Using Graph Neural Networks

Introduction of the Coordination Polyhedron Graph Network (CPGN), a multi-scale GNN for more accurate crystal property prediction in materials science.

Hardware/Chips arXiv cs.AI

Dynamic Multi-Criteria Bottleneck Severity Index (DMBSI) for Semiconductor Wafer Manufacturing: A Genetically Optimised Framework for Reentrant Production Systems

A data-driven framework (DMBSI) for identifying and analyzing bottleneck severity in semiconductor wafer manufacturing process flows.

AI/ML arXiv cs.AI

Foundation Models for EEG Are Blind to Long-Range Temporal Correlations: A Spectral-Temporal Dissociation Behind Their Cross-Population Fragility

Study revealing that current EEG foundation models fail to capture long-range temporal correlations, affecting their robustness across populations.

AI/ML arXiv cs.AI

MedJudgeRAG: Option-Wise Evidence Judgment with Dynamic Knowledge Graphs for Medical MCQA

MedJudgeRAG uses dynamic knowledge graphs to improve medical multiple-choice question answering by better judging evidence from retrieved documents.

AI/ML arXiv cs.AI

REPREC: Representation Driven Parameter-Efficient Recommendation System

REPREC is a lightweight framework for sequential recommendation that aligns user representations without needing to fine-tune the LLM backbone.

AI/ML arXiv cs.AI

Two Views, One Voice: Evidence-Grounded Conversational Music Recommendation

A decoupled retrieval and response pipeline for conversational music recommendation that enhances explanation reliability using evidence-grounded generation.

AI/ML arXiv cs.AI

Extremal Chowla sets and their linear analogues: A human-AI mathematical investigation using Co-Scientist

A mathematical investigation into extremal Chowla sets and their linear analogues, developed through collaboration between humans and the Co-Scientist AI tool.

Software Engineering Hacker News

A Texture Lookup Approach to Bézier Curve Evaluation on the GPU (JCGT)

A technical exploration of using texture lookup to evaluate Bezier curves on GPUs, likely improving rendering performance.

Cybersecurity Hacker News

Disrupting supply chain attacks on NPM and GitHub Actions

Research on disrupting supply chain attacks targeting the NPM ecosystem and GitHub Actions pipelines.

Software Engineering Hacker News

ALP: Adaptive lossless floating-point compression

Introduction of ALP, a new method for adaptive lossless floating-point compression.