AI

AgentAuditor, developed by researchers from multiple universities, enables AI evaluators to reach human-level precision in assessing the safety and security of intelligent agents.

Google releases Gemma 3n, a compact, high-performance multimodal model with 10 billion parameters that runs on just 2GB RAM, advancing edge AI capabilities for mobile devices.

Tsinghua NLP Lab introduces RLPR, a novel reinforcement learning method that enhances reasoning across diverse fields, overcoming previous limitations in natural language understanding.

Ant Group unveils 'AQ', an AI healthcare app offering over 100 functions including health info, diagnosis, report interpretation, and hospital connections, now available across major app stores.

Five large AI models participated in the 2025 Shandong college entrance exam, with Gemini and Doubao achieving top scores in science and literature, showcasing rapid AI progress in academic assessments.

ICCV 2025 will be held in Hawaii with a 24% acceptance rate from over 11,200 submissions, reflecting rapid growth in computer vision research and increasing conference competitiveness.

Researchers from Peking University and ByteDance Seed introduce μP theory to optimize hyperparameters, enabling scalable and efficient diffusion transformer models with minimal search costs.

Harvard researchers developed ultra-thin, flexible bioelectronic devices that seamlessly integrate with developing vertebrate brains, enabling long-term neural recording without damage.

DeepMind's AlphaGenome analyzes up to 1 trillion DNA bases, predicting gene regulation and mutations at single-base resolution, advancing understanding of genetic diseases.

Meta has aggressively recruited top AI researchers from OpenAI's Zurich office, including three ViT authors, amid ongoing competition in AI talent acquisition, as reported by WSJ.