Individuals with a birth weight of less than one kilogram have an increased risk of cerebral palsy, intellectual disabilities and attention deficit disorders. Nonetheless, they rate their quality of life as high as individuals with normal birth weight. This finding emerges from a new study of 201 adults who were born prematurely and followed from birth to the age of 26–29 in Sweden.

Researchers from King Abdullah University of Science and Technology (KAUST) have developed deepBlastoid, the first deep-learning platform specifically designed for the high-throughput, automated classification of human stem cell-derived embryo models (blastoids). By leveraging a ResNet-18 architecture and a novel Confidence Rate metric, the model achieves up to 97% accuracy and processes images 1,000 times faster than human experts. This tool facilitates large-scale drug screening and basic research into early human development by providing a standardized, objective evaluation framework.

New cutting-edge research, led by Academy Professor Otso Ovaskainen of the University of Jyväskylä and David Dunson at Duke University, combines citizen bird observations with artificial intelligence and the computing power of supercomputers at CSC – IT Center for Science. The international and multidisciplinary research team has developed the world’s most accurate prediction model, capable of anticipating even small shifts in bird occurrence almost in real time. 

Scientists have introduced the first absolute, contactless thermometry technique based on Brillouin scattering in gases, enabling high-accuracy temperature measurements from room temperature down to deep cryogenic levels. By filling hollow-core fibres with gases such as neon, argon, or helium, the method retrieves temperature directly from the Brillouin frequency shift via the interaction between laser light and acoustic waves, eliminating the need for calibration. The results pave the way for ultra-precise, minimally invasive thermometry in extreme environments.

For the first time, researchers have experimentally discovered optical meron-antimeron pairs—a novel class of topological quasiparticles that were originally conceived in particle physics and later found in magnets, but had never been realized in optics. By employing group representation theory, the team not only generated these states in deep-subwavelength plasmonic systems but also established a complete symmetry classification that dictates their topological invariants. This framework reveals a fundamental chirality-parity locking effect and enables the creation of isolated merons. Harnessing the robustness of these symmetry-protected states, the work demonstrates an ultra-sensitive sensing application with a record λ/133 resolution, bridging abstract theory to practical photonic technology.

Researchers from the Faculty of Engineering at The University of Hong Kong (HKU) have developed two innovative deep-learning algorithms, ClairS-TO and Clair3-RNA, that significantly advance genetic mutation detection in cancer diagnostics and RNA-based genomic studies.

Talking to oneself is a trait which feels inherently human - but it’s not just humans who can reap the benefits of such self-talk. Published in Neural Computation, scientists from the Okinawa Institute of Science and Technology (OIST) have demonstrated the potential of inner speech to improve AI learning, showing how AI models can generalize across different tasks more easily when supported by both inner speech and short-term memory.