Enregistré dans:
Détails bibliographiques
Auteurs principaux: Je, Gyung Hyun, Raffel, Colin
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2512.24991
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866911348228620288
author Je, Gyung Hyun
Raffel, Colin
author_facet Je, Gyung Hyun
Raffel, Colin
contents While large language models (LLMs) demonstrate reasonable zero-shot capability across many downstream tasks, fine-tuning is a common practice to improve their performance. However, a task's data efficiency--i.e., the number of fine-tuning examples needed to achieve a desired level of performance--is often unknown, resulting in costly cycles of incremental annotation and retraining. Indeed, we demonstrate across a curated set of 30 specialized tasks that performant LLMs may struggle zero-shot but can attain stronger performance after fine-tuning. This motivates the need for methods to predict a task's data efficiency without requiring incremental annotation. After introducing a concrete metric that quantifies a task's data efficiency, we propose using the gradient cosine similarity of low-confidence examples to predict data efficiency based on a small number of labeled samples. We validate our approach on a diverse set of tasks with varying data efficiencies, attaining 8.6% error in overall data efficiency prediction and typically eliminating hundreds of unnecessary annotations on each task. Our experiment results and implementation code are available on GitHub.
format Preprint
id arxiv_https___arxiv_org_abs_2512_24991
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Efficiently Estimating Data Efficiency for Language Model Fine-tuning
Je, Gyung Hyun
Raffel, Colin
Machine Learning
While large language models (LLMs) demonstrate reasonable zero-shot capability across many downstream tasks, fine-tuning is a common practice to improve their performance. However, a task's data efficiency--i.e., the number of fine-tuning examples needed to achieve a desired level of performance--is often unknown, resulting in costly cycles of incremental annotation and retraining. Indeed, we demonstrate across a curated set of 30 specialized tasks that performant LLMs may struggle zero-shot but can attain stronger performance after fine-tuning. This motivates the need for methods to predict a task's data efficiency without requiring incremental annotation. After introducing a concrete metric that quantifies a task's data efficiency, we propose using the gradient cosine similarity of low-confidence examples to predict data efficiency based on a small number of labeled samples. We validate our approach on a diverse set of tasks with varying data efficiencies, attaining 8.6% error in overall data efficiency prediction and typically eliminating hundreds of unnecessary annotations on each task. Our experiment results and implementation code are available on GitHub.
title Efficiently Estimating Data Efficiency for Language Model Fine-tuning
topic Machine Learning
url https://arxiv.org/abs/2512.24991