Featured video: Coding for underwater robotics
Lincoln Laboratory intern Ivy Mahncke developed and tested algorithms to help human divers and robots navigate underwater.
New method could increase LLM training efficiency
By leveraging idle computing time, researchers can double the speed of model training while preserving accuracy.
AI to help researchers see the bigger picture in cell biology
By providing holistic information on a cell, an AI-driven method could help scientists better understand disease mechanisms and plan experiments.
Exposing biases, moods, personalities, and abstract concepts hidden in large language models
A new method developed at MIT could root out vulnerabilities and improve LLM safety and performance.
Parking-aware navigation system could prevent frustration and emissions
By minimizing the need to drive around looking for a parking spot, this technique can save drivers up to 35 minutes — and give them a realistic estimate of total travel time.
AI algorithm enables tracking of vital white matter pathways
Opening a new window on the brainstem, a new tool reliably and finely resolves distinct nerve bundles in live diffusion MRI scans, revealing signs of injury or disease.
Helping AI agents search to get the best results out of large language models
EnCompass executes AI agent programs by backtracking and making multiple attempts, finding the best set of outputs generated by an LLM. It could help coders work with AI agents more efficiently.
MIT engineers design structures that compute with heat
By leveraging excess heat instead of electricity, microscopic silicon structures could enable more energy-efficient thermal sensing and signal processing.
3 Questions: How AI could optimize the power grid
While the growing energy demands of AI are worrying, some techniques can also help make power grids cleaner and more efficient.
One pull of a string is all it takes to deploy these complex structures
A new method could enable users to design portable medical devices, like a splint, that can be rapidly converted from flat panels to a 3D object without any tools.
Guided learning lets “untrainable” neural networks realize their potential
CSAIL researchers find even “untrainable” neural nets can learn effectively when guided by another network’s built-in biases using their guidance method.
A new way to increase the capabilities of large language models
MIT-IBM Watson AI Lab researchers developed an expressive architecture that provides better state tracking and sequential reasoning in LLMs over long texts.
Working to eliminate barriers to adopting nuclear energy
Nuclear waste continues to be a bottleneck in the widespread use of nuclear energy, so doctoral student Dauren Sarsenbayev is developing models to address the problem.
Enabling small language models to solve complex reasoning tasks
The “self-steering” DisCIPL system directs small models to work together on tasks with constraints, like itinerary planning and budgeting.
MIT affiliates named 2025 Schmidt Sciences AI2050 Fellows
Postdoc Zongyi Li, Associate Professor Tess Smidt, and seven additional alumni will be supported in the development of AI against difficult problems.