AI

Recent research exposes significant security flaws in AI-powered computer assistants, highlighting risks like phishing, malicious commands, and privacy breaches, emphasizing the need for safety benchmarks.

Researchers from Tsinghua and other institutions enhance protein folding generation by integrating global geometric perception into diffusion models, advancing structural coverage and interpretability.

This article explores the differences between how deep neural networks and humans perceive the world, focusing on whether they rely on pixels or abstract concepts for recognition and understanding.

Meta has aggressively recruited top AI talent, including 8 Chinese researchers from prestigious universities, forming a powerful team to advance its AI research and development efforts.

Researchers from Berkeley and Meta develop a world model for embodied intelligence, allowing AI to predict visual outcomes based on full-body actions, advancing human-like perception and planning.

Quark, Alibaba's AI flagship application, has generated over 10 million professional volunteer reports, marking the largest deep research application in China, driven by advanced task planning technology.

Meta has officially established the 'Superintelligence Laboratory,' assembling an elite team of 11 top AI experts from leading institutions to spearhead future AI innovations.

A joint publication by UofT, UBC, MIT, and Fudan offers a systematic review of diffusion models in anomaly detection and generation, highlighting recent advances and future directions.

Young AI talents born after 1995 are simultaneously transforming industries and publishing top conference papers, showcasing rapid growth and innovative spirit in the AI field.

A novel hierarchical reasoning model with just 27 million parameters achieves superior performance over DeepSeek and Claude, demonstrating efficient, interpretable reasoning with minimal data.