AI/ML arXiv cs.AI

ReasonCLIP-58M: Visually Grounded Commonsense Reasoning Supervision for CLIP

ReasonCLIP-58M is a framework for integrating large-scale reasoning supervision into CLIP-style models to improve visually grounded commonsense reasoning.

AI/ML arXiv cs.AI

NaviCache: Test-Time Self-Calibration Caching for Video Generation

NaviCache is a test-time self-calibration method to reduce computational costs in video generation by skipping redundant computations.

AI/ML Hacker News

Show HN: Smart model routing directly in Claude, Codex and Cursor

A new tool enabling smart model routing across Claude, Codex, and Cursor for optimized AI model usage.

Other TechCrunch

Tesla settles FSD crash lawsuit as federal investigations continue

Tesla settles a lawsuit involving a fatal crash linked to its Full Self-Driving system as federal probes continue.

Tech Business/VC TechCrunch

TikTok’s road to becoming a super app

TikTok is reportedly expanding its functionality to evolve into a comprehensive super app for various digital activities.

AI/ML The Verge

OpenAI unveils GPT-5.6 amid US AI regulatory drama

OpenAI introduces GPT-5.6, a suite of models (Sol, Terra, and Luna) focused on coding, cybersecurity, and biology, with competitive pricing.

Tech Business/VC The Verge

Smart lock maker Level has been gutted and its founders are out

Assa Abloy has laid off most of Level Home staff and is folding the smart lock business into Kwikset.

Other Ars Technica

Ars Live: What's the latest in the aftermath of the New Glenn catastrophe?

A live stream discussion regarding the aftermath of the New Glenn rocket catastrophe.

AI/ML arXiv cs.AI

MLFFM-SegDiff: A Multi-Level Feature Fusion Diffusion Model for Skin Lesion Segmentation

Research presents MLFFM-SegDiff, a multi-level feature fusion diffusion model designed to improve skin lesion segmentation accuracy.

AI/ML arXiv cs.AI

Robust Onion: Peeling Open Vocab Object Detectors Under Noise

The 'Robust Onion' study analyzes the impact of noise on Open Vocabulary Object Detectors, proposing a lightweight plug-and-play approach for better robustness.

AI/ML arXiv cs.AI

Anatomy-Guided Residual Motion Diffusion for Controllable 4D Cardiac MRI Synthesis

A new framework for controllable 4D cardiac MRI synthesis using latent diffusion models to improve medical image data augmentation.

AI/ML arXiv cs.AI

AIGP: An LLM-Based Framework for Long-Term Value Alignment in E-Commerce Pricing

AIGP is an LLM-based framework that aligns e-commerce pricing decisions with long-term business value using offline reinforcement learning.

Software Engineering Hacker News

Run isolated sandboxes with full lifecycle control: AWS introduces MicroVMs

AWS introduces MicroVMs to allow developers to run isolated sandboxes with full lifecycle control.

Tech Business/VC TechCrunch

It’s not about Anthropic vs. OpenAI anymore

An opinion piece discussing the political consequences of advancing AI models and the need for collective action.

AI/ML arXiv cs.AI

CAT-Q: Cost-efficient and Accurate Ternary Quantization for LLMs

Introduction of CAT-Q, a cost-efficient post-training ternary quantization scheme for LLMs that significantly reduces training token requirements.

AI/ML arXiv cs.AI

LAMP: Lane-Aligned Motion Primitives for Feasible Trajectory Prediction

LAMP is a topology-aware forecasting framework for autonomous driving that uses VQ-VAE to predict lane-aligned motion primitives.

AI/ML arXiv cs.AI

Zero-Shot Size Transfer for Neural ODEs on Sparse Random Graphs: Graphon Limits and Adjoint Convergence

Theoretical research on zero-shot size transfer for Neural ODEs on sparse random graphs using graphon limits.

Cybersecurity arXiv cs.AI

TGHE: Template-based Graph Homomorphic Encryption for Privacy-Preserving GNN Inference in Edge-Cloud Systems

TGHE provides a template-based graph homomorphic encryption framework to speed up privacy-preserving GNN inference in edge-cloud systems.

AI/ML arXiv cs.AI

Disco-LoRA: Disentangled Composition of Content, Style, and Motion for Multi-concept Video Customization

Disco-LoRA is a unified framework for multi-concept video customization, disentangling content, style, and motion using a dual-LoRA approach.

AI/ML arXiv cs.AI

Beyond Logical Forms: LLM-Extracted Patterns for Fallacy Classification

A study on using LLMs to extract patterns for better automated classification of logical fallacies.