Researchers use statistical physics and "toy models" to explain how neural networks avoid overfitting and stabilize learning in high-dimensional spaces.

The quest for more training data has created a glut of low-quality junk data that could derail the promise of physical AI.

WORK Medical continues to deepen its commitment to the medical AI sector. This partnership represents the latest advancement of the Company’s “Healthcare + Web3 + AI” strategy, particularly in the ...

Net, a hybrid model that improves energy consumption prediction in low-energy buildings, enhancing accuracy and ...

EVOLVE, an agentic framework that autonomously optimizes AI training data, model architectures, and learning algorithms — boosting MMLU scores by 18 points over human baselines.

Artificial intelligence systems based on neural networks—such as ChatGPT, Claude, DeepSeek or Gemini—are extraordinarily ...

Many drug and antibody discovery pathways focus on intricately folded cell membrane proteins. When molecules of a drug ...

The promise of smart test is a data-chain problem before it is an algorithm problem. A device can pass every checkpoint and ...

Database increased by 60% from approximately 25,000 meters of drilling to over 40,000 meters; Development, integration and refinement of the model will continue with new data obta ...

ERNIE 5.1 hits the top of Chinese AI leaderboards while spending a fraction of what rivals do. Baidu calls it a "parameter ...

For years, machines have navigated the world color-blind. LiDAR sensors – the laser-based eyes of self-driving cars, ...