A smarter way for large language models to think about hard problems
Reports and Proceedings
Updates every hour. Last Updated: 20-Apr-2026 22:16 ET (21-Apr-2026 02:16 GMT/UTC)
MIT researchers developed a smarter way for an LLM to allocate computation while it reasons about a problem. Their method lets the model dynamically adjust its computational budget based on the difficulty of the question and the likelihood that each partial solution will lead to the correct answer.
MIT researchers developed an aerial microrobot that can fly with speed and agility comparable to real insects. The research opens the door to future bug-sized robots that could aid in search-and-rescue missions.
As disruptions to GPS services increase globally, radio signals from low-Earth orbit (LEO) satellites could become reliable navigation alternatives, a new study suggests.
HER2-targeted therapy showed durable and clinically meaningful results for patients with rare bile duct cancers
Researchers identified promising therapeutic targets for KRAS-mutant cancers and kidney cancers
Study provides insights into AI models for genomic analysis and clinical decision-making
Researchers uncovered how DNA shape and rigidity affects gene expression
LMU researchers conduct representative survey on the perception and use of AI in journalism.