KAIST-MIT-Microsoft Develop AI Technology to Restore Compressed Visual Information to High Resolution Without Additional Training

GPU (Graphics Processing Unit) memory efficiency improved by up to 16 times... expected to accelerate commercialization of humanoid robots and on-device AI

Following paper acceptance at 'CVPR 2026', the world's most prestigious conference in artificial intelligence, global technological prowess and research reliability were proven by winning the 'CVPR Compute Gold Star' and being selected as 'Transparency Champion'

Inspired by the human brain, Oregon State University researchers have developed a new light-sensitive device that combines sensing and memory while controlling how digital memories strengthen or fade over time.

Idiopathic pulmonary fibrosis remains a hard-to-treat lung disease with limited effective drugs. A recent study in Engineering used machine learning to screen natural compounds and found dihydromyricetin, a flavonoid from herbs, can block the TGF-β/ALK5 signaling pathway. It alleviates fibrosis and inflammation in cell and mouse models, shows good safety, and offers a new natural-product-based direction for pulmonary fibrosis treatment.

ORNL tested two aluminum alloys — cast ACMZ (AlCuMnZr) and printed DuAlumin3D — in a prototype GM engine. The materials enabled a more-than-10-percent improvement in fuel efficiency and a 15-percent reduction in weight without sacrificing strength or durability. 

Based on the underlying paper, the key innovation is not simply better toxicity detection, but an adaptive cascade of AI models. Instead of sending every post through a large, computationally expensive detector, the system starts with faster, lightweight classifiers. If a piece of content is clearly benign or clearly toxic, a decision is made immediately. Only ambiguous cases are passed to increasingly powerful models. A reinforcement-learning algorithm decides which model to consult next, balancing speed, accuracy and computing cost. This approach dramatically increases throughput while slightly improving detection accuracy. (ScienceDirect)

For a 50–60 word EurekAlert summary:

Concordia researchers have developed an AI system that detects toxic online content faster and more accurately by combining multiple detection models in an adaptive sequence. Simple cases are screened by lightweight classifiers, while only difficult content is sent to more powerful AI tools. The approach improves accuracy while processing content up to nine times faster than conventional methods.