30
36氪首发 | 英伟达全球首家触觉仿真合作伙伴再获数亿融资，上半年订单超去年四倍
Chinese AI tactile sensing startup Tashan Technology has secured hundreds of millions of yuan in a Series B funding round to advance its sensor and chip technologies.
29
SensorFM: Towards a general intelligence and interface for wearable health data
Google Research has introduced SensorFM, a foundation model designed to act as a general intelligence interface for analyzing wearable health data.
29
Cyborg Cockroach Gets 3D-Printed Diving Suit: Flood Rescue Breakthrough Using Biomimetic Insects
Researchers from NTU Singapore and Waseda University have developed a 3D-printed diving suit for cyborg cockroaches to assist in underwater search and rescue operations.
29
OpenAI bets on families as ChatGPT goes deeper into households
OpenAI is hiring a dedicated product manager to develop ChatGPT features tailored for families, caregivers, and older adults.
29
Shares soar as recently-listed Zhipu AI and Iluvatar CoreX launch secondary placements
Chinese artificial intelligence developer Zhipu AI and chipmaker Iluvatar CoreX have launched secondary share placements in Hong Kong to raise billions of dollars for hardware and research.
29
AI华语歌，终于能听了！从零预训练十亿参数，告别「人机味」
ByteDance has entered the AI music generation market by pre-training a one-billion-parameter model focused on producing high-quality Chinese vocal music.
29
全球首个获工业级全域CE认证的具身智能机器人！无界动力K15开启全球批量交付
Boundless Power has announced that its K15 embodied AI robot has achieved EU CE certification and begun global commercial deliveries.
28
Modular by necessity: Why the AI boom is making prefabricated data centers the new default
The rapid demand for AI compute is driving the industry toward prefabricated, modular data centers to bypass traditional multi-year construction timelines.
28
FansAI收购OPC公司新映科技
AI content and interactive entertainment company FansAI has fully acquired the AI music application developer Xinying Technology.
28
Instagram’s Adam Mosseri: If you don’t like AI, ‘then you shouldn’t have it in your feed’
Instagram head Adam Mosseri stated that the platform should label AI-generated content rather than filter it out entirely, suggesting users customize their feeds if they dislike it.
28
Minimal Decision Dynamics and Contextual Probability: A Quantum Tug-of-War Model
Researchers developed a quantum-like extension of the Tug-of-War model to represent context-dependent decision-making using a minimal internal state.
28
Interval Certifications for Multilayered Perceptrons via Lattice Traversal
This paper proposes a theoretical framework that reduces the adversarial robustness problem of multilayered perceptrons to a lattice traversal problem.
28
Ceci n'est pas une pipe: AI systems as semantic abstractions
The authors propose a semantic framework to evaluate the correctness of AI outputs by distinguishing between domain knowledge, reference sources, and system capabilities.
28
赛力斯：预计上半年净亏损15亿元-18亿元
Chinese electric vehicle manufacturer Seres expects a net loss of 1.5 billion to 1.8 billion yuan for the first half of the year, citing rising raw material costs and asset adjustments.
28
How did the government decide OpenAI’s frontier model was safe to release?
Questions remain regarding the specific nature of the safety discussions between the government and AI labs like OpenAI and Anthropic prior to model releases.
28
Global AI Market Shift: Why Developers Are Choosing Cost-Efficient Models
Developers are increasingly adopting cost-efficient AI models from Chinese providers like DeepSeek, Zhipu AI, and Qwen, which now account for over 20 percent of OpenRouter's weekly token volume.
27
GPT-5.5 Bio Bug Bounty
OpenAI has launched a biological safety bug bounty program focused on identifying potential biosecurity risks associated with its GPT-5.5 model.
27
对话北森CEO纪伟国：账上躺着16亿现金，AI转型去往何方？ | SaaS+Agent十人谈
Beisen CEO Ji Weiguo discussed the company's transition toward AI-native solutions, highlighted by the launch of their new AI HR expert platform, Mavens.
27
Large Tabular Models Excel Where LLMs Fail
Researchers are developing specialized large tabular models to address the limitations of traditional large language models in analyzing structured data.
27
UltraX: Refining Pre-Training Data at Scale with Adaptive Programmatic Editing
UltraX is an adaptive programmatic editing framework designed to refine large-scale pre-training datasets efficiently to improve language model performance.
27
When the Judge Changes, So Does the Measurement: Auditing LLM-as-Judge Reliability
An empirical audit of LLM-as-judge systems shows that upgrading or changing the evaluator model leads to inconsistent scoring, highlighting reliability challenges in automated evaluation.
27
把35B模型塞进32GB内存，智能体PC如何挑战端侧部署的「物理极限」？
Intel and its partners showcased several applications demonstrating the feasibility of running a 35-billion parameter model on consumer PCs with 32GB of RAM.
27
ByteDance Seedream 5.0 Pro: The Strongest Image Model Returns With Precision Editing and Multilingual Generation
ByteDance's Seed team has released Seedream 5.0 Pro, an image generation model featuring enhanced editing capabilities and multilingual support across multiple scenarios.
27
Open source AI matters more than ever, according to Hugging Face’s Clem Delangue
Hugging Face CEO Clem Delangue emphasized the growing adoption of open-source AI models and datasets among major corporations.