{"id":9500,"date":"2026-07-27T12:30:29","date_gmt":"2026-07-27T10:30:29","guid":{"rendered":"https:\/\/www.kubicek.ai\/?post_type=lexicon&#038;p=9500"},"modified":"2026-07-27T12:57:08","modified_gmt":"2026-07-27T10:57:08","slug":"overfitting","status":"publish","type":"lexicon","link":"https:\/\/www.kubicek.ai\/en\/lexicon\/overfitting\/","title":{"rendered":"Overfitting"},"content":{"rendered":"<p class=\"wp-block-paragraph\"><strong>Overfitting<\/strong> is a state in which a machine-learning model achieves a very low error on the training data but performs markedly worse on data it has never seen. Instead of the general pattern, the model has memorized the specific samples \u2013 noise, outliers and random artefacts of data collection included. Formally it is an unfavourable point on the bias-variance tradeoff: the capacity of the model is too high relative to the amount and diversity of the data. It is diagnosed from the widening gap between the training and validation error curves \u2013 the training error keeps falling while the validation error levels off and begins to rise. The standard defences are regularization (L1 and L2 weight penalties, dropout in the hidden layers), early stopping, augmentation and expansion of the dataset, cross-validation and simplification of the architecture. In large language models, overfitting also shows up as verbatim memorization of passages from the training corpus, which is a security and a copyright problem at the same time.<\/p>\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n<p class=\"wp-block-paragraph\">Picture a student who memorizes twenty specific past exam papers before the test \u2013 including the fact that the right answer happened to be &#8220;c&#8221; three times in a row. On a test made up of exactly those questions he scores a hundred percent. The moment the examiner changes the numbers or the wording, he falls apart, because he never grasped the principle. That is precisely what an overfitted model does: brilliant on what it has seen, helpless on everything else. The opposite is the student who studied too superficially and cannot even manage the old papers \u2013 that is called underfitting. The aim of training is to find the point between these two extremes.<\/p>\n","protected":false},"featured_media":0,"template":"","class_list":["post-9500","lexicon","type-lexicon","status-publish","hentry"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.kubicek.ai\/en\/wp-json\/wp\/v2\/lexicon\/9500","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kubicek.ai\/en\/wp-json\/wp\/v2\/lexicon"}],"about":[{"href":"https:\/\/www.kubicek.ai\/en\/wp-json\/wp\/v2\/types\/lexicon"}],"wp:attachment":[{"href":"https:\/\/www.kubicek.ai\/en\/wp-json\/wp\/v2\/media?parent=9500"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}