3
ShinyHunters: How to close the Credential Rotation Gap
An analysis explores the technical challenges and failure points associated with maintaining secure credential rotation across organizations.
3
基金二季报科技调仓现分歧，多只绩优基金维持高仓位
Public fund managers are showing divergent strategies regarding technology sector investments despite the ongoing global AI boom.
3
ByteDance denies plans to enter smart driving business
ByteDance has refuted media reports claiming that the company plans to enter the autonomous driving and driverless logistics sectors through its cloud platform, Volcano Engine.
3
机构：HBM4价格今年下半年或涨至4-5美元/千比特
High-bandwidth memory HBM4 prices are projected to rise significantly in the second half of 2026 due to surging AI demand and manufacturing complexities.
3
科技行业裁员浪潮加剧，多数美国劳动者支持设立人工智能财富基金
A recent survey reveals that a majority of Americans support legislation to mandate that major AI companies contribute half of their shares to a public sovereign wealth fund to distribute AI dividends.
3
Roblox launches an AI-powered game-creation feature in its mobile app
Roblox has integrated a text-to-game generation feature into its mobile application to simplify the game creation process.
3
Telecom Italia board approves Poste Italiane takeover bid
Telecom Italia's board has approved a takeover bid that will return the company to state ownership.
3
Spatiotemporal Facial Action Unit Detection using Twin Cycle Autoencoders for Driver Monitoring
A twin cycle autoencoder approach is proposed for spatiotemporal facial action unit detection to enhance driver monitoring systems under challenging conditions.
3
RealDESED: A Real-World Domestic Sound Event Detection Benchmark
Researchers present RealDESED, a real-world audio benchmark dataset collected from homes to improve domestic sound event detection models.
3
MaxSAT-Based Feedback for Guiding Vision-Language Models in Sudoku
Researchers propose a neuro-symbolic approach that uses formal constraint reasoning to improve the logical consistency of vision-language models on Sudoku puzzles.
3
The Aspirations Of HPE And Dell In The Quantum-Classical HPC Datacenter
Hewlett Packard Enterprise and Dell are outlining their strategic goals for integrating quantum and classical high-performance computing within modern datacenters.
3
长飞光纤等在武汉成立智能创投基金，出资额5亿
Yangtze Optical Fibre and Cable has partnered with other entities to establish a 500 million yuan venture capital fund in Wuhan focused on smart technologies.
3
Valfortec plans €300m data center campus in Alicante, Spain
Valfortec is planning to develop a 60MW data center campus in Alicante, Spain, with an investment of 300 million euros.
3
How Smartsheet built a remote MCP server on AWS
Smartsheet has detailed the architecture and AWS-based infrastructure used to deploy a remote Model Context Protocol server.
3
SPQR: A Multi-Dimensional Benchmark for Safety Alignment under Benign Model Adaptation
A new benchmark, SPQR, evaluates the stability of safety alignment in text-to-image models when subjected to common downstream fine-tuning techniques.
3
Reason Less, Verify More: Deterministic Gates Recover a Silent Policy-Violation Failure Mode in Tool-Using LLM Agents
Researchers identify a failure mode where LLM agents successfully complete tasks while violating safety policies, and propose deterministic gates as a mitigation strategy.
3
Agent Hacks Agent: Autoresearch for Production-Agent Red-Teaming
Researchers propose a new red-teaming framework that uses autonomous agents to identify security vulnerabilities in production LLM agent workflows.
3
To Answer or to Abstain: Mitigating Search-Agent Hallucinations via Abstention-Aware Reinforcement Learning
Researchers have developed an abstention-aware reinforcement learning approach to reduce hallucinations in search-augmented LLMs by penalizing incorrect answers when retrieval fails.
3
Remembering Distinct Items, Not Tokens: A Learnable Dirichlet-Process Cache Between State-Space Models and Attention
This paper proposes a learnable Dirichlet-process cache that stores only novel inputs to bridge the gap between state-space models and attention mechanisms.
3
高德发布通用世界模型工坊ABot-World Studio：5090单卡可生成小时级实时交互式视频与3D场景
AutoNavi has launched ABot-World Studio, a tool that enables the generation of interactive 3D scenes and videos using a single consumer-grade GPU.
3
HCRMap: Pressure-Aware Hot-Expert Residency Mapping for 3.5D MoE Chiplet Inference
The paper introduces HCRMap, a mapping framework designed to mitigate compute and communication imbalances caused by uneven expert activation in Mixture-of-Experts models on multi-chiplet systems.
3
MemDecay: Region-Aware KV Cache Eviction for Efficient LLM Agent Inference
MemDecay is a new memory management technique that optimizes LLM agent inference by using region-aware KV cache eviction based on semantic structure.
3
ActiveFly-Bench: Aligning Embodied Question Answering with Vision-Language-Action for Aerial Embodied Perception
ActiveFly-Bench is a new benchmark designed to evaluate aerial embodied perception by linking high-level task understanding, behavior planning, and low-level control for unmanned aerial vehicles.
3
Unlocking Parallelism in Autoregressive Language Models via Speculative Decoding with Progressive Tree Drafting
Researchers propose a new speculative decoding method that improves large language model inference speed by utilizing progressive tree drafting to better exploit parallel processing.