Understanding and Controlling Repetition Neurons and Induction Heads in In-Context Learning

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Doan, Nhi Hoai, Hiraoka, Tatsuya, Inui, Kentaro
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915611479638016
author Doan, Nhi Hoai
Hiraoka, Tatsuya
Inui, Kentaro
author_facet Doan, Nhi Hoai
Hiraoka, Tatsuya
Inui, Kentaro
contents This paper investigates the relationship between large language models' (LLMs) ability to recognize repetitive input patterns and their performance on in-context learning (ICL). In contrast to prior work that has primarily focused on attention heads, we examine this relationship from the perspective of skill neurons, specifically repetition neurons. Our experiments reveal that the impact of these neurons on ICL performance varies depending on the depth of the layer in which they reside. By comparing the effects of repetition neurons and induction heads, we further identify strategies for reducing repetitive outputs while maintaining strong ICL capabilities.
format Preprint
id arxiv_https___arxiv_org_abs_2507_07810
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Understanding and Controlling Repetition Neurons and Induction Heads in In-Context Learning
Doan, Nhi Hoai
Hiraoka, Tatsuya
Inui, Kentaro
Computation and Language
This paper investigates the relationship between large language models' (LLMs) ability to recognize repetitive input patterns and their performance on in-context learning (ICL). In contrast to prior work that has primarily focused on attention heads, we examine this relationship from the perspective of skill neurons, specifically repetition neurons. Our experiments reveal that the impact of these neurons on ICL performance varies depending on the depth of the layer in which they reside. By comparing the effects of repetition neurons and induction heads, we further identify strategies for reducing repetitive outputs while maintaining strong ICL capabilities.
title Understanding and Controlling Repetition Neurons and Induction Heads in In-Context Learning
topic Computation and Language
url https://arxiv.org/abs/2507.07810