Homelab/Self-Hosting Hacker News

Announcing .self: A New Top-Level Domain Designed to Support Self-Hosting

Introduction of .self, a new top-level domain specifically designed to facilitate and encourage self-hosting.

Open Source Hacker News

ACL 1.0: A source-available commercial license for the AI era

Announcement of ACL 1.0, a source-available commercial license tailored for the AI era.

AI/ML TechCrunch

Gemini’s personalized AI image generation is now free for U.S. users

Google makes personalized AI image generation in Gemini free for eligible users in the U.S.

Cybersecurity The Verge

Leaked iPhone 18 Pro photos reportedly wound up on the dark web

Leaked iPhone 18 Pro photos and parts lists appear on the dark web following a supplier data breach.

AI/ML arXiv cs.AI

iCost: A Novel Instance-Complexity-Based Cost-Sensitive Learning Framework

Introduction of iCost, an instance-complexity-based cost-sensitive learning framework for imbalanced classification, released as a scikit-learn compatible package.

AI/ML arXiv cs.AI

Deepfake Media Generation and Detection in the Generative AI Era: A Survey and Outlook

A comprehensive survey on deepfake media generation and detection, including a new multimodal benchmark for evaluating detectors.

AI/ML arXiv cs.AI

Derivation of effective gradient flow equations and dynamical truncation of training data in Deep Learning

Research deriving gradient flow equations for deep learning with ReLU activations to improve the interpretability of supervised learning.

AI/ML arXiv cs.AI

The Minimal Search Space for Conditional Causal Bandits

A study on conditional causal bandits proposing an efficient algorithm to identify the minimal search space for optimal conditional interventions.

AI/ML arXiv cs.AI

ReFreeKV: Towards Threshold-Free KV Cache Compression

ReFreeKV is a threshold-free KV cache compression method for LLMs that adaptively adjusts budget allocation to maintain performance.

AI/ML arXiv cs.AI

DMind Benchmark: Toward a Holistic Assessment of LLM Capabilities across the Web3 Domain

The DMind Benchmark is a comprehensive evaluation suite for assessing LLM capabilities within the Web3 domain, focusing on security and reasoning.

Open Source Hacker News

JumpServer: Open-Source Privileged Access Management

JumpServer is an open-source Privileged Access Management (PAM) solution designed to secure and audit access to servers and network devices.

Cybersecurity Hacker News

.garden TLD's change to a bad neighborhood

The .garden TLD is increasingly associated with malicious activity and spam, becoming a 'bad neighborhood' for web users.

AI/ML Hacker News

Micro-Agent: Beat Frontier Models with Collaboration Inside Model API

Micro-Agent is a technique that uses collaboration within a model API to allow smaller models to outperform larger frontier models.

Tech Business/VC The Verge

Tidal won’t pay royalties on AI-generated music but isn’t banning it outright

Tidal has announced it will label 100% AI-generated music and will stop paying royalties to such tracks to prioritize human artists.

AI/ML arXiv cs.AI

Edit-R2: Context-Aware Reinforcement Learning for Multi-Turn Image Editing

Edit-R2 is a reinforcement learning framework that improves multi-turn image editing by reconstructing session intent to maintain consistency across iterations.

AI/ML arXiv cs.AI

RedKnot: Efficient Long-Context LLM Serving with Head-Aware KV Reuse and SegPagedAttention

RedKnot is a head-aware KV cache management system that optimizes LLM serving efficiency by decomposing the KV cache along attention heads.

AI/ML arXiv cs.AI

Auto-Configuring Scientific Simulators with Lightweight Coding-Agent Adapters

SIGA is a lightweight coding-agent adapter that simplifies the configuration of scientific simulators by providing the necessary executable contracts.

AI/ML arXiv cs.AI

Agent-as-a-Router: Agentic Model Routing for Coding Tasks

The Agent-as-a-Router (ACRouter) framework optimizes model routing for coding tasks using a feedback loop to reduce cumulative regret.

AI/ML arXiv cs.AI

POTracker: Optimizing Large Language Models for Standard-Compliant Power Outage Report Generation

POTracker is an optimized LLM for generating standard-compliant power outage reports using a new loss function for structural accuracy.

Cybersecurity arXiv cs.AI

AI Snitches Get Glitches: Towards Evading Agentic Surveillance

Researchers introduce SurveilBench to evaluate the surveillance capabilities of AI agents and propose prompt injection techniques to evade such surveillance.