Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Meyer, Maxime, Michelessa, Mario, Chaux, Caroline, Tan, Vincent Y. F.
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2509.00421
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911130554728448
author Meyer, Maxime
Michelessa, Mario
Chaux, Caroline
Tan, Vincent Y. F.
author_facet Meyer, Maxime
Michelessa, Mario
Chaux, Caroline
Tan, Vincent Y. F.
contents Despite the empirical success of prompt tuning in adapting pretrained language models to new tasks, theoretical analyses of its capabilities remain limited. Existing theoretical work primarily addresses universal approximation properties, demonstrating results comparable to standard weight tuning. In this paper, we explore a different aspect of the theory of transformers: the memorization capability of prompt tuning. We provide two principal theoretical contributions. First, we prove that the amount of information memorized by a transformer cannot scale faster than linearly with the prompt length. Second, and more importantly, we present the first formal proof of a phenomenon empirically observed in large language models: performance degradation in transformers with extended contexts. We rigorously demonstrate that transformers inherently have limited memory, constraining the amount of information they can retain, regardless of the context size. This finding offers a fundamental understanding of the intrinsic limitations of transformer architectures, particularly their ability to handle long sequences.
format Preprint
id arxiv_https___arxiv_org_abs_2509_00421
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Memory Limitations of Prompt Tuning in Transformers
Meyer, Maxime
Michelessa, Mario
Chaux, Caroline
Tan, Vincent Y. F.
Machine Learning
Despite the empirical success of prompt tuning in adapting pretrained language models to new tasks, theoretical analyses of its capabilities remain limited. Existing theoretical work primarily addresses universal approximation properties, demonstrating results comparable to standard weight tuning. In this paper, we explore a different aspect of the theory of transformers: the memorization capability of prompt tuning. We provide two principal theoretical contributions. First, we prove that the amount of information memorized by a transformer cannot scale faster than linearly with the prompt length. Second, and more importantly, we present the first formal proof of a phenomenon empirically observed in large language models: performance degradation in transformers with extended contexts. We rigorously demonstrate that transformers inherently have limited memory, constraining the amount of information they can retain, regardless of the context size. This finding offers a fundamental understanding of the intrinsic limitations of transformer architectures, particularly their ability to handle long sequences.
title Memory Limitations of Prompt Tuning in Transformers
topic Machine Learning
url https://arxiv.org/abs/2509.00421