AI/ML arXiv cs.AI

From Intent to Infrastructure: LLM-Driven Agent Compilers for ISAC Networks

The Agent Compiler uses LLMs to translate high-level engineering intent into executable ISAC system configurations for 6G networks.

AI/ML arXiv cs.AI

Physics-Informed Feature Engineering 1D-CNN for Multilayer Cloud Detection from Geostationary Satellites

A physics-informed 1D-CNN is developed to improve multilayer cloud detection from geostationary satellites using radiative transfer theory.

AI/ML arXiv cs.AI

ForensicNet: Lightweight Attention-Enhanced MobileNetV2 for Automated Face Identification

ForensicNet is a lightweight, attention-enhanced MobileNetV2 framework designed for real-time automated face identification in forensic settings.

Other Hacker News

The Great Steam Deck Crash of 2026: How the New Pricing Decreased Demand by 80%

A discussion regarding the impact of price increases on the demand for the Steam Deck in 2026.

Homelab/Self-Hosting Hacker News

Motion Sensors and Home Security Gadgets Without Cameras

A conversation about motion sensors and home security devices that prioritize privacy by avoiding cameras.

Tech Business/VC TechCrunch

Bluecore Energy raises $10M to build portable nuclear reactors on barges

Bluecore Energy has raised $10 million in pre-seed funding to develop portable nuclear reactors on barges.

Other The Verge

The Switch 2 is $50 off at Woot for new customers

Woot is offering discounts on the Nintendo Switch 2 before a planned price increase in September 2026.

Hardware/Chips The Verge

The wireless headset that gets hot-swappable batteries right

The Glorious GHS InfinitePlay wireless headset features a high-efficiency hot-swappable battery system with zero audio downtime.

Other The Verge

Nintendo keeps making shooters more friendly and approachable

Analysis of Nintendo's approach to making competitive shooters more accessible, specifically through the Splatoon series.

Hardware/Chips The Verge

Halliday’s latest smart glasses feature a much-improved display

Halliday's Gen 2 smart glasses offer an improved display experience compared to the finicky first generation.

AI/ML arXiv cs.AI

Feature Generation Using LLMs: An Evolutionary Algorithm Approach

Researchers propose an evolutionary algorithm approach using LLMs to automate feature generation for tabular data in ML pipelines.

AI/ML arXiv cs.AI

Discovery by Dreaming: Cross-Domain Recombination in Artificial Memory

A study introduces 'Discovery by Dreaming,' using cross-domain recombination in artificial memory to enable AI discovery and creative insight.

AI/ML arXiv cs.AI

From Outcomes to Actions: Leveraging Hindsight for Long-Horizon Language Agent Training

Introduction of Hindsight Policy Optimization (HPO), a novel policy gradient method to improve long-horizon language agent training using intent space projection.

AI/ML arXiv cs.AI

TRACE: Trajectory-Based Safety Patch Learning for LLM Post-Training Realignment

Introduces TRACE, a framework for learning safety patches to realign LLMs after customized fine-tuning without destroying task-specific utility.

AI/ML arXiv cs.AI

RobustMAD: Evaluating Real-World Robustness of Multimodal Small Language Models for Deployable Anomaly Detection Assistants

Presents RobustMAD, a benchmark for evaluating the real-world robustness of multimodal small language models used in industrial anomaly detection.

AI/ML arXiv cs.AI

CIGPO: Contextual Information-Gain Policy Optimization for Multi-Turn Evidence-Reading LLM Agents

Proposes CIGPO, a policy optimization method using contextual information-gain rewards to prevent reward-variance collapse in multi-turn LLM agents.

Hardware/Chips arXiv cs.AI

Learning Structural Manipulability in Gate-Level Netlists Using Graph Neural Networks

Uses Graph Neural Networks to learn structural manipulability in gate-level netlists, providing insights into circuit topology and Trojan detection.

AI/ML arXiv cs.AI

Let the Data Decide: Supervision Analysis, Capability Trade-offs, and Adaptive Objective Routing in Continued Pre-Training via Off-Policy Distillation

Analyzes the trade-offs between language-modeling and knowledge-distillation objectives during LLM continued pre-training via off-policy distillation.

AI/ML arXiv cs.AI

High-accuracy Low-Bit KV-Cache Quantization via Local Distribution Restoration

Introduces DGAP, a local distribution restoration technique to significantly improve the accuracy of low-bit KV-cache quantization in LLMs.

Other arXiv cs.AI

An Agentic Interface for End-to-End Probabilistic Seismic Hazard and Risk Analysis

Develops an agentic interface for seismic hazard and risk analysis using the Model Context Protocol (MCP) to wrap the OpenQuake engine.