Pressemitteilung 11/2025 - Bayerisches Staatsministerium der Justiz

From Memorization to Reasoning in the Spectrum of Loss Curvature

We characterize how memorization is represented in transformer models and show that it can be disentangled in the weights of both language models (LMs) and vision transformers (ViTs) using a decomposition based on the loss landscape curvature. This insight is based on prior theoretical and empirical work showing that the curvature for memorized training points is much sharper than non memorized, meaning ordering weight components from high to low curvature can reveal a distinction without explicit labels. This motivates a weight editing procedure that suppresses far more recitation of untargeted memorized data more effectively than a recent unlearning method (BalancedSubnet), while maintaining lower perplexity. Since the basis of curvature has a natural interpretation for shared structure in model weights, we analyze the editing procedure extensively on its effect on downstream tasks in LMs, and find that fact retrieval and arithmetic are specifically and consistently negatively affected, even though open book fact retrieval and general logical reasoning is conserved. We posit these tasks rely heavily on specialized directions in weight space rather than general purpose mechanisms, regardless of whether those individual datapoints are memorized. We support this by showing a correspondence between task data's activation strength with low curvature components that we edit out, and the drop in task performance after the edit. Our work enhances the understanding of memorization in neural networks with practical applications towards removing it, and provides evidence for idiosyncratic, narrowly-used structures involved in solving tasks like math and fact retrieval.

Slow and Steady, Kay Ryan’s “Turtle” Poem Will Win Your Heart

A.O. Scott ponders the specific gravity and unlikely grace of Kay Ryan’s “Turtle.” And we have a game to help you memorize it.

How much information do LLMs really memorize? Now we know, thanks to Meta, Google, Nvidia and Cornell

Using a clever solution, researchers find GPT-style models have a fixed memorization capacity of approximately 3.6 bits per parameter.

【生成AIパスポート試験対策】 GPTs - エミリーと学ぶ生成AIの世界

「【生成AIパスポート試験対策】 GPTs」 のご紹介です。 この試験に関する生成AI動画が、AIを…