Propulsion: Steering LLM with Tiny Fine-Tuning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kowsher, Md, Prottasha, Nusrat Jahan, Bhat, Prakash
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915064912543744
author Kowsher, Md
Prottasha, Nusrat Jahan
Bhat, Prakash
author_facet Kowsher, Md
Prottasha, Nusrat Jahan
Bhat, Prakash
contents The rapid advancements in Large Language Models (LLMs) have revolutionized natural language processing (NLP) and related fields. However, fine-tuning these models for specific tasks remains computationally expensive and risks degrading pre-learned features. To address these challenges, we propose Propulsion, a novel parameter efficient fine-tuning (PEFT) method designed to optimize task-specific performance while drastically reducing computational overhead. Inspired by the concept of controlled adjustments in physical motion, Propulsion selectively re-scales specific dimensions of a pre-trained model, guiding output predictions toward task objectives without modifying the model's parameters. By introducing lightweight, trainable Propulsion parameters at the pre-trained layer, we minimize the number of parameters updated during fine-tuning, preventing overfitting or overwriting of existing knowledge. Our theoretical analysis, supported by Neural Tangent Kernel (NTK) theory, shows that Propulsion approximates the performance of full fine-tuning with far fewer trainable parameters. Empirically, Propulsion reduces the parameter count from 355.3 million to just 0.086 million, achieving over a 10x reduction compared to standard approaches like LoRA while maintaining competitive performance across benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2409_10927
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Propulsion: Steering LLM with Tiny Fine-Tuning
Kowsher, Md
Prottasha, Nusrat Jahan
Bhat, Prakash
Computation and Language
The rapid advancements in Large Language Models (LLMs) have revolutionized natural language processing (NLP) and related fields. However, fine-tuning these models for specific tasks remains computationally expensive and risks degrading pre-learned features. To address these challenges, we propose Propulsion, a novel parameter efficient fine-tuning (PEFT) method designed to optimize task-specific performance while drastically reducing computational overhead. Inspired by the concept of controlled adjustments in physical motion, Propulsion selectively re-scales specific dimensions of a pre-trained model, guiding output predictions toward task objectives without modifying the model's parameters. By introducing lightweight, trainable Propulsion parameters at the pre-trained layer, we minimize the number of parameters updated during fine-tuning, preventing overfitting or overwriting of existing knowledge. Our theoretical analysis, supported by Neural Tangent Kernel (NTK) theory, shows that Propulsion approximates the performance of full fine-tuning with far fewer trainable parameters. Empirically, Propulsion reduces the parameter count from 355.3 million to just 0.086 million, achieving over a 10x reduction compared to standard approaches like LoRA while maintaining competitive performance across benchmarks.
title Propulsion: Steering LLM with Tiny Fine-Tuning
topic Computation and Language
url https://arxiv.org/abs/2409.10927