Researchers found a better way to teach large language models new skills
Peer-Reviewed Publication
Updates every hour. Last Updated: 21-Aug-2025 07:10 ET (21-Aug-2025 11:10 GMT/UTC)
Researchers have developed a technique that significantly improves the performance of large language models without increasing the computational power necessary to fine-tune the models. The researchers demonstrated that their technique improves the performance of these models over previous techniques in tasks including commonsense reasoning, arithmetic reasoning, instruction following, code generation, and visual recognition.
Instead of a tempest in a teapot, imagine the cosmos in a canister. Scientists have performed experiments using nested, spinning cylinders to confirm that an uneven wobble in a ring of electrically conductive fluid like liquid metal or plasma causes particles on the inside of the ring to drift inward. Since revolving rings of plasma also occur around stars and black holes, these new findings imply that the wobbles can cause matter in those rings to fall toward the central mass and form planets.
The scientists found that the wobble could grow in a new, unexpected way. Researchers already knew that wobbles could grow from the interaction between plasma and magnetic fields in a gravitational field. But these new results show that wobbles can more easily arise in a region between two jets of fluid with different velocities, an area known as a free shear layer.
Researchers from the Faculty of Physics at the University of Warsaw and the University of British Columbia have described how a so-called lone spinon - an exotic quantum excitation that is a single unpaired spin - can arise in magnetic models. The discovery deepens our understanding of the nature of magnetism and could have implications for the development of future technologies such as quantum computers and new magnetic materials. The findings were published in the renowned journal “Physical Review Letters”.
The intricate, hidden processes that sustain coral life are being revealed through a new microscope developed by scientists at UC San Diego’s Scripps Institution of Oceanography.
The diver-operated microscope — called the Benthic Underwater Microscope imaging PAM, or BUMP — incorporates pulse amplitude modulated (PAM) light techniques to offer an unprecedented look at coral photosynthesis on micro-scales. Funded by the National Science Foundation, the new microscope will help scientists uncover precisely why corals bleach, and inform remediation efforts. While the bleaching process is known, it’s not fully understood, and it hasn’t been possible to study in depth in the field — until now.
Researchers have long been interested in how humans and animals make decisions by focusing on trial-and-error behavior informed by recent information. However, the conventional frameworks for understanding these behaviors may overlook certain realities of decision-making because they assume we make the best decisions after taking into account our past experiences. A newly released study by a team of scientists deploys AI in innovative ways to better understand this process. By using tiny artificial neural networks, the researchers’ work illuminates in detail what drives an individual’s actual choices—regardless of whether those choices are optimal or not.