Tiny traps, big trouble: Small regions within cells aggregate proteins linked to ALS, dementia
Peer-Reviewed Publication
Updates every hour. Last Updated: 17-Jan-2026 07:11 ET (17-Jan-2026 12:11 GMT/UTC)
MIT researchers find that large language models sometimes mistakenly link certain grammatical sequences to specific topics, and then rely on these learned patterns when answering queries. This phenomenon can cause LLMs to fail unexpectedly on new tasks and could be exploited by adversarial agents to trick an LLM into generating harmful content.
Florida State University oceanographers have discovered a significant connection between small-scale microbial processes and ecosystem-wide dynamics, offering new insights into the mechanisms driving marine carbon storage.
Lehigh University researcher Hannah Dailey is leading a new international collaboration to improve predictions of how bone fractures heal. Supported by the U.S. National Science Foundation and the Swiss National Science Foundation, the four-year project partners with the AO Research Institute Davos (ARI) to develop computational models that combine mechanical factors—such as implant stiffness and loading patterns—with biological processes that vary from patient to patient. Using ARI’s extensive imaging library documenting fracture healing in sheep, the team will build probabilistic models capable of forecasting how recovery will progress. The models will ultimately be integrated into ARI’s online training platform to help surgeons understand how implant choices and rehabilitation strategies influence healing. Long term, the goal is to enable patient-specific simulations that help clinicians identify complications earlier and make more informed treatment decisions.
Researchers at the University of Cincinnati found that monk parakeets introduced to new birds will “test the waters” with potential friends to avoid increasingly dangerous close encounters that could lead to injury.