{"id":203540,"date":"2026-04-09T09:00:00","date_gmt":"2026-04-09T13:00:00","guid":{"rendered":"https:\/\/testing.news-you-need.com\/index.php\/2026\/04\/09\/new-technique-makes-ai-models-leaner-and-faster-while-theyre-still-learning-mit-news\/"},"modified":"2026-04-09T09:18:25","modified_gmt":"2026-04-09T13:18:25","slug":"new-technique-makes-ai-models-leaner-and-faster-while-theyre-still-learning-mit-news","status":"publish","type":"post","link":"https:\/\/testing.news-you-need.com\/index.php\/2026\/04\/09\/new-technique-makes-ai-models-leaner-and-faster-while-theyre-still-learning-mit-news\/","title":{"rendered":"New technique makes AI models leaner and faster while they\u2019re still learning | MIT News"},"content":{"rendered":"<p><a href=\"https:\/\/news.mit.edu\/2026\/new-technique-makes-ai-models-leaner-faster-while-still-learning-0409\">New technique makes AI models leaner and faster while they\u2019re still learning | MIT News<\/a><\/p>\n<p><a href=\"https:\/\/news.mit.edu\/2026\/new-technique-makes-ai-models-leaner-faster-while-still-learning-0409\">https:\/\/news.mit.edu\/2026\/new-technique-makes-ai-models-leaner-faster-while-still-learning-0409<\/a><\/p>\n<p>Publish Date: <a href=\"publish_date]\">2026-04-09 09:00:00<\/a><\/p>\n<p>Source Domain: <a href=\"news.mit.edu\">news.mit.edu<\/a><\/p>\n<ul>\n<li>\n<p><strong>New Compression Method During Training:<\/strong> Researchers have developed CompreSSM, a technique to compress state-space models during training, avoiding the trade-off between model size and performance improvement.<\/p>\n<\/li>\n<li>\n<p><strong>Advantages Over Traditional Methods:<\/strong> Unlike conventional pruning or knowledge distillation methods, CompreSSM enables informed compression mid-training, saving computational resources and time without sacrificing accuracy.<\/p>\n<\/li>\n<li>\n<p><strong>Efficiency and Performance:<\/strong> On benchmarks like CIFAR-10 and Mamba, CompreSSM-trained models match the performance of larger models while training up to 1.5 times faster and achieving up to 4x training speedups.<\/p>\n<\/li>\n<li>\n<p><strong>Theoretical Grounding:<\/strong> The method leverages control theory, specifically Hankel singular values, to rank model components by their importance early in training, ensuring only significant parts are retained.<\/p>\n<\/li>\n<li>\n<p><strong>Safe Compression:<\/strong> CompreSSM includes a safety net feature allowing practitioners to revert to saved checkpoints if a compression step adversely affects performance.<\/p>\n<\/li>\n<li>\n<p><strong>Effectiveness on Certain Models:<\/strong> The method works particularly well on multi-input, multi-output models and linear time-invariant systems, with potential extensions to other state-space model architectures, including transformers.<\/p>\n<\/li>\n<li>\n<p><strong>Future Directions:<\/strong> Researchers aim to extend CompreSSM to more complex architectures and dynamical systems, aiming for wider applicability in modern AI systems.<\/p>\n<\/li>\n<li>\n<p><strong>Conference and Support:<\/strong> The work was accepted for presentation at the International Conference on Learning Representations 2026, supported by several institutions including the Max Planck Institute and U.S. Office of Naval Research.<\/p>\n<\/li>\n<\/ul>\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>New technique makes AI models leaner and faster while they\u2019re still learning | MIT News&#8230;<\/p>\n","protected":false},"author":1,"featured_media":203541,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"https:\/\/news.mit.edu\/sites\/default\/files\/images\/202603\/mit-csail-compressm.jpg","fifu_image_alt":"","footnotes":""},"categories":[14],"tags":[],"class_list":["post-203540","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-artificial-intelligence"],"_links":{"self":[{"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/posts\/203540"}],"collection":[{"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/comments?post=203540"}],"version-history":[{"count":1,"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/posts\/203540\/revisions"}],"predecessor-version":[{"id":203542,"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/posts\/203540\/revisions\/203542"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/media\/203541"}],"wp:attachment":[{"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/media?parent=203540"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/categories?post=203540"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/tags?post=203540"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}