{"id":204890,"date":"2026-04-22T15:15:00","date_gmt":"2026-04-22T19:15:00","guid":{"rendered":"https:\/\/testing.news-you-need.com\/index.php\/2026\/04\/22\/teaching-ai-models-to-say-im-not-sure-mit-news\/"},"modified":"2026-04-22T15:35:16","modified_gmt":"2026-04-22T19:35:16","slug":"teaching-ai-models-to-say-im-not-sure-mit-news","status":"publish","type":"post","link":"https:\/\/testing.news-you-need.com\/index.php\/2026\/04\/22\/teaching-ai-models-to-say-im-not-sure-mit-news\/","title":{"rendered":"Teaching AI models to say \u201cI\u2019m not sure\u201d | MIT News"},"content":{"rendered":"<p><a href=\"https:\/\/news.mit.edu\/2026\/teaching-ai-models-to-say-im-not-sure-0422\">Teaching AI models to say \u201cI\u2019m not sure\u201d | MIT News<\/a><\/p>\n<p><a href=\"https:\/\/news.mit.edu\/2026\/teaching-ai-models-to-say-im-not-sure-0422\">https:\/\/news.mit.edu\/2026\/teaching-ai-models-to-say-im-not-sure-0422<\/a><\/p>\n<p>Publish Date: <a href=\"publish_date]\">2026-04-22 15:15:00<\/a><\/p>\n<p>Source Domain: <a href=\"news.mit.edu\">news.mit.edu<\/a><\/p>\n<ul>\n<li>\n<p><strong>Overconfidence in AI Systems:<\/strong> Modern AI reasoning models, such as those at MIT&#8217;s CSAIL, express answers with the same high level of certainty regardless of whether they are right or guessing, a problem traced to their training methods.<\/p>\n<\/li>\n<li>\n<p><strong>Issue with Reinforcement Learning:<\/strong> The training method for these models, which rewards only correctness without considering correctness by chance, fosters overconfidence, leading to unreliable outputs in critical applications.<\/p>\n<\/li>\n<li>\n<p><strong>RLCR Method Developed:<\/strong> Researchers have introduced RLCR (Reinforcement Learning with Calibration Rewards), a technique that trains models to output both answers and calibrated confidence estimates, effectively addressing overconfidence. <\/p>\n<\/li>\n<li>\n<p><strong>Effective Results:<\/strong> RLCR reduced calibration errors by up to 90% in experiments while either maintaining or improving accuracy on both trained and new tasks.<\/p>\n<\/li>\n<li>\n<p><strong>Practical Utility:<\/strong> The confidence estimates generated by RLCR improve both the accuracy and calibration when used for selecting or weighting candidate answers.<\/p>\n<\/li>\n<li>\n<p><strong>Added Value of Uncertainty Reasoning:<\/strong> Including a model\u2019s uncertainty reasoning in its input data enhanced classifier performance, indicating that self-awareness about uncertainty holds practical value.<\/p>\n<\/li>\n<\/ul>\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Teaching AI models to say \u201cI\u2019m not sure\u201d | MIT News https:\/\/news.mit.edu\/2026\/teaching-ai-models-to-say-im-not-sure-0422 Publish Date: 2026-04-22&#8230;<\/p>\n","protected":false},"author":1,"featured_media":204891,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"https:\/\/news.mit.edu\/sites\/default\/files\/images\/202604\/mit-csail-reinforcement.jpg","fifu_image_alt":"","footnotes":""},"categories":[14],"tags":[],"class_list":["post-204890","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\/204890"}],"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=204890"}],"version-history":[{"count":1,"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/posts\/204890\/revisions"}],"predecessor-version":[{"id":204892,"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/posts\/204890\/revisions\/204892"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/media\/204891"}],"wp:attachment":[{"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/media?parent=204890"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/categories?post=204890"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/testing.news-you-need.com\/index.php\/wp-json\/wp\/v2\/tags?post=204890"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}