AI/ML arXiv cs.AI

Training Therapeutic Judges and Multi-Agent Systems for Human-Aligned Mental Health Support

Introduction of TheraJudge and TheraAgent, a framework for improving mental health LLM responses through human-aligned evaluation and refinement.

AI/ML arXiv cs.AI

Curvature-Guided Module Localization for Low-Rank Detoxification of Backdoored Large Language Models

A weight-space repair framework using curvature-guided module localization to detoxify backdoored LLMs without full retraining.

AI/ML arXiv cs.AI

How Human Feedback Shapes AI-generated Community Notes

Analysis of human feedback patterns in AI-generated Community Notes on social media platforms like X and Facebook.

AI/ML arXiv cs.AI

Budget-Adaptive Routing: Skipping the Weak When the Strong Answers Anyway

Proposed budget-adaptive routing for edge-cloud inference that reduces latency by skipping weak detectors when a strong model is likely to answer.

AI/ML arXiv cs.AI

Behavior Cloning is Not All You Need: The Optimality of On-Policy Distillation for Noisy Expert Feedback

Theoretical study proving that on-policy distillation outperforms supervised fine-tuning when learning from noisy expert feedback in LLMs.

AI/ML arXiv cs.AI

Physics-informed Conditional Normalizing Flows for Angles-only Cislunar Orbit Determination

Application of physics-informed conditional normalizing flows for improving orbit determination in cislunar environments.

Other Hacker News

SpudCell: The first synthetic cell with a complete cell cycle

Discussion regarding SpudCell, the first synthetic cell capable of completing a full cell cycle.

Other Hacker News

The Anti-Palantir Manifesto

A manifesto detailing a critical philosophical and technical stance against the practices of Palantir.

Other Hacker News

1-Bit Pixel Art Emojis

A technical exploration or showcase of 1-bit pixel art emojis.

Other Hacker News

Internal Combustion Engine

Technical discussion regarding the internal combustion engine.

Other Hacker News

This Cell Feeds, Grows and Reproduces. and It's Manmade

News about a manmade cell that can feed, grow, and reproduce, marking a milestone in synthetic biology.

Tech Business/VC The Verge

Comcast’s split could make or break Peacock

Analysis of how Comcast's corporate split may impact the viability of its streaming service, Peacock.

Cybersecurity arXiv cs.AI

AI-Generated PowerShell Malware: An Experimental Framework and Dataset

Research presenting an experimental framework and dataset to analyze PowerShell malware generated by LLMs.

AI/ML arXiv cs.AI

A Stationary-Distribution Theory for Triplet-Based Plateau Search in Random Forest Ensemble-Size Selection

A theoretical study on stationary-distribution for ensemble-size selection in Random Forest models using plateau search.

AI/ML arXiv cs.AI

Test-Time Verification for Text-to-SQL via Outcome Reward Models

Introduction of GradeSQL, a framework that uses Outcome Reward Models (ORMs) to improve Text-to-SQL reliability via test-time verification.

AI/ML arXiv cs.AI

The Label Imitation Game: Turing Test Network for Zero-Shot Pseudo-Label Pruning

Introduction of the Label Imitation Game (LIG) and Turing Test Network (TTN) for zero-shot pseudo-label pruning in vision-language models.

Other Hacker News

For First Time, a Cell Built from Scratch Grows and Divides

Researchers have successfully created a synthetic cell from scratch that is capable of growing and dividing, marking a major milestone in synthetic biology.

Other Hacker News

Sony Deletes 551 Movies PlayStation Owners Paid For

Sony has reportedly deleted hundreds of movies that users had paid for, raising concerns about digital ownership and content preservation.

Tech Business/VC Hacker News

Monetization Gateway

A discussion regarding a 'Monetization Gateway', though the content provided is limited to comments.

Other The Verge

The best Switch 2 case I’ve tried is cheaper than usual

Review and deals for a Nintendo Switch 2 case and gaming monitors, focused on consumer electronics accessories.