Saved in:
Bibliographic Details
Main Author: Lin, Dongyan
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2406.18926
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916303729590272
author Lin, Dongyan
author_facet Lin, Dongyan
contents Fine-tuning pretrained language models has shown promising results on a wide range of tasks, but when encountering a novel task, do they rely more on generic pretrained representation, or develop brand new task-specific solutions? Here, we fine-tuned GPT-2 on a context-dependent decision-making task, novel to the model but adapted from neuroscience literature. We compared its performance and internal mechanisms to a version of GPT-2 trained from scratch on the same task. Our results show that fine-tuned models depend heavily on pretrained representations, particularly in later layers, while models trained from scratch develop different, more task-specific mechanisms. These findings highlight the advantages and limitations of pretraining for task generalization and underscore the need for further investigation into the mechanisms underpinning task-specific fine-tuning in LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2406_18926
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Fine-tuned network relies on generic representation to solve unseen cognitive task
Lin, Dongyan
Machine Learning
Fine-tuning pretrained language models has shown promising results on a wide range of tasks, but when encountering a novel task, do they rely more on generic pretrained representation, or develop brand new task-specific solutions? Here, we fine-tuned GPT-2 on a context-dependent decision-making task, novel to the model but adapted from neuroscience literature. We compared its performance and internal mechanisms to a version of GPT-2 trained from scratch on the same task. Our results show that fine-tuned models depend heavily on pretrained representations, particularly in later layers, while models trained from scratch develop different, more task-specific mechanisms. These findings highlight the advantages and limitations of pretraining for task generalization and underscore the need for further investigation into the mechanisms underpinning task-specific fine-tuning in LLMs.
title Fine-tuned network relies on generic representation to solve unseen cognitive task
topic Machine Learning
url https://arxiv.org/abs/2406.18926