New technique makes AI models leaner and faster while they’re still learning

Researchers use control theory to shed unnecessary complexity from AI models during training, cutting compute costs without sacrificing performance. Rachel Gordon | MIT CSAIL Publication Date: April 9, 2026 Press Inquiries Press Contact: Rachel Gordon Email: rachelg@csail.mit.edu Phone: 617-258-0675 MIT Computer Science and Artificial Intelligence Laboratory Close Caption: A new technique, called CompreSSM, helps identify which parts of a model are pulling their weight before surgically removing unnecessary components early in the training process. Credits: Image: Alex Shipps/MIT CSAIL and Makram Chahine, using assets from Pixabay and Pexels. Previous image Next image Training a large artificial intelligence model is expensive, not just in dollars, but in time, energy, and computational resources. Traditionally, obtaining a smaller, faster model either requires training a massive one first and then trimming it down, or training a small one from scratch and accepting weaker performance. Researchers at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL), Max Planck Institute for Intelligent Systems, European Laboratory for Learning and Intelligent Systems, ETH, and Liquid AI have now developed a new method that sidesteps this trade-off entirely, compressing models during training, rather than after.The technique, called CompreSSM, targets a family of AI architectures known as state-space models, which power applications ranging from language processing to a...