Saved in:
Bibliographic Details
Main Authors: Deb, Rohan, Thekumparampil, Kiran, Kalantari, Kousha, Hiranandani, Gaurush, Sabach, Shoham, Kveton, Branislav
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.14826
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912385216806912
author Deb, Rohan
Thekumparampil, Kiran
Kalantari, Kousha
Hiranandani, Gaurush
Sabach, Shoham
Kveton, Branislav
author_facet Deb, Rohan
Thekumparampil, Kiran
Kalantari, Kousha
Hiranandani, Gaurush
Sabach, Shoham
Kveton, Branislav
contents Supervised fine-tuning (SFT) is a standard approach to adapting large language models (LLMs) to new domains. In this work, we improve the statistical efficiency of SFT by selecting an informative subset of training examples. Specifically, for a fixed budget of training examples, which determines the computational cost of fine-tuning, we determine the most informative ones. The key idea in our method is to select examples that maximize information gain, measured by the Hessian of the log-likelihood of the LLM. We approximate it efficiently by linearizing the LLM at the last layer using multinomial logistic regression models. Our approach is computationally efficient, analyzable, and performs well empirically. We demonstrate this on several problems, and back our claims with both quantitative results and an LLM evaluation.
format Preprint
id arxiv_https___arxiv_org_abs_2505_14826
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle FisherSFT: Data-Efficient Supervised Fine-Tuning of Language Models Using Information Gain
Deb, Rohan
Thekumparampil, Kiran
Kalantari, Kousha
Hiranandani, Gaurush
Sabach, Shoham
Kveton, Branislav
Machine Learning
Computation and Language
Supervised fine-tuning (SFT) is a standard approach to adapting large language models (LLMs) to new domains. In this work, we improve the statistical efficiency of SFT by selecting an informative subset of training examples. Specifically, for a fixed budget of training examples, which determines the computational cost of fine-tuning, we determine the most informative ones. The key idea in our method is to select examples that maximize information gain, measured by the Hessian of the log-likelihood of the LLM. We approximate it efficiently by linearizing the LLM at the last layer using multinomial logistic regression models. Our approach is computationally efficient, analyzable, and performs well empirically. We demonstrate this on several problems, and back our claims with both quantitative results and an LLM evaluation.
title FisherSFT: Data-Efficient Supervised Fine-Tuning of Language Models Using Information Gain
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2505.14826