2
When your brain works differently, AI isn’t a luxury—it’s accessibility
An author describes how AI-powered desktop assistants can serve as crucial accessibility tools to support neurodivergent professionals with executive functioning.
2
SpaceX falls to $135 IPO price ahead of Starship launch
Market valuation for SpaceX has declined following its initial public offering as investor expectations adjust to the company's operational performance.
2
Monitor Amazon SageMaker Pipelines cross-account with custom Amazon CloudWatch dashboards
AWS has released a new monitoring solution that allows users to track SageMaker pipelines across multiple accounts and regions using CloudWatch dashboards.
2
5 Trends That Defined AI Engineering at World’s Fair 2026
The 2026 AI Engineering World's Fair highlighted a shift toward designing complex systems around AI agents rather than simply utilizing individual agents.
2
威兆半导体递表港交所
Semiconductor manufacturer WeiZhao Semiconductor has officially filed for an initial public offering on the Hong Kong Stock Exchange.
2
Dexmal Aims to Ignite Embodied AI Productivity With DM0.5 Model, DexOS, and MaaS Three-Stage Strategy
Dexmal has launched its DM0.5 foundation model, Apex universal robot, and DexOS operating system to bridge the gap between embodied AI models and real-world industrial productivity.
2
CSC Financial: Semiconductor Equipment Supply Chain Faces Dual Localization Trend and Major Investment Opportunity
CSC Financial forecasts that the semiconductor equipment market will grow by 23.5% in 2026 to reach $152.2 billion, driven by an AI-fueled cycle and localization trends.
2
TopoExplore: Topological Discrimination for Archive-Based Exploration
TopoExplore is a new reinforcement learning exploration method that uses topological analysis to determine if unexplored regions are physically accessible.
2
SMETA-ZSL:Semantic Meta-Alignment for Zero-Shot Threat Classification
The paper proposes SMETA-ZSL, a zero-shot learning framework that leverages large language models to classify emerging cybersecurity threats without labeled training data.
2
腾讯等入股语用科技
Tencent and HongShan have invested in Yuyong Technology, a newly established software development and data processing services company.
2
Verifier-Guided Twelve-Tone Composition: A Generate-Verify-Repair Harness for Symbolic Music Generation
A new neuro-symbolic framework improves the consistency of LLM-generated twelve-tone music by implementing a generate-verify-repair loop.
2
Vertu wants executives to pay $6,880 for an AI agent — here’s how it actually performs
A review of a high-end luxury smartphone highlights the practical performance and limitations of its integrated AI agent features.
2
Mitigating Early Training Collapse in CTR Models
This study analyzes the phenomenon of early training collapse in click-through rate prediction models and evaluates practical mitigation strategies using large-scale industrial datasets.
2
Comparing Socially-Equitable Renewable Energy Budget Allocation MDP Policies in Mature and Emerging Economies
This study models socially equitable renewable energy budget allocation as a Markov Decision Process to compare policy strategies between mature and emerging economies.
2
Personalized Emotional Intelligence in Generative AI through Symbolic Affective Reasoning
Researchers introduced a symbolic affective reasoning framework to help generative AI models better understand, predict, and personalize emotional responses.
2
GRC-ProbNet: Uncertainty-aware Feature Extraction for Cardiovascular Disease Classification
GRC-ProbNet introduces uncertainty-aware feature extraction to improve the classification of cardiovascular disease from CT images by accounting for segmentation mask variability.
2
GRATE: Temporal Extensions for Inductive KG Foundation Models via Gated Rotary Attention
Researchers developed GRATE, a model that extends inductive transfer capabilities to temporal knowledge graphs using gated rotary attention.
2
CLIR-Bench: Benchmarking Multimodal Question Answering over Irregular Clinical Time Series
The authors presented CLIR-Bench, a benchmark designed to evaluate multimodal question answering models on sparse and irregularly sampled clinical time-series data.
2
Refused in Chat, Written in Code: Workflow-Level Jailbreak Construction in IDE Coding Agents
Researchers demonstrate a new security vulnerability where coding agents can be manipulated into performing harmful tasks through multi-stage, workflow-level jailbreak attacks.
2
MultiView-Bench: A Diagnostic Benchmark for World-Centric Multi-View Integration in VLMs
MultiView-Bench is a new diagnostic benchmark designed to evaluate the ability of vision-language models to integrate information across multiple viewpoints for 3D scene comprehension.
2
Seeing is Free, Speaking is Not: Uncovering the True Energy Bottleneck in Edge VLM Inference
Systematic energy profiling of edge-based vision-language models reveals that language processing, rather than visual token processing, is the primary driver of power consumption.
2
Signed Symmetric Quantization for Few-Bit Integers
This work demonstrates that clipping positive outliers in standard symmetric integer quantization causes significant errors at few-bit precision and proposes an alternative approach.
2
Sticky Routing: Training MoE Models for Memory-Efficient Inference
Researchers have introduced StickyMoE, a routing consistency loss designed to reduce weight swapping and improve memory efficiency during Mixture-of-Experts model inference on edge devices.
2
Deceptive Grounding: Entity Attribution Failure in Clinical Retrieval-Augmented Generation
Researchers identify a critical failure mode in clinical retrieval-augmented generation systems where models correctly cite sources but misattribute clinical evidence to the wrong medical entities.