RIFF: Learning to Rephrase Inputs for Few-shot Fine-tuning of Language Models

Saeed Najafi, Alona Fyshe
3/4/2024
cs.CLcs.LG

Official code: SaeedNajafi/RIFF (0 stars, MIT, last commit 2024-08-20).

Abstract

Pre-trained Language Models (PLMs) can be accurately fine-tuned for downstream text processing tasks. Recently, researchers have introduced several parameter-efficient fine-tuning methods that optimize input prompts or adjust a small number of model parameters (e.g LoRA). In this study, we explore the impact of altering the input text of the original task in conjunction with parameter-efficient fine-tuning methods. To most effectively rewrite the input text, we train a few-shot paraphrase model with a Maximum-Marginal Likelihood objective. Using six few-shot text classification datasets, we show that enriching data with paraphrases at train and test time enhances the performance beyond what can be achieved with parameter-efficient fine-tuning alone. The code used for our experiments can be found at https://github.com/SaeedNajafi/RIFF.

View on arXivView PDF

Code Implementations(1)

SaeedNajafi/RIFFOfficial100%
MIT

Learning to Rephrase Inputs for Few-shot FineTuning

00Shell, PythonMay 27, 20232 years agoMIT

Similar Papers

Small Language Models Improve Giants by Rewriting Their Outputs

84% match1 repo122023

LoPT: Low-Rank Prompt Tuning for Parameter Efficient Language Models

84% match2024

Parameter-Efficient Fine-Tuning of Large Language Models using Semantic Knowledge Tuning

84% match2024

FPT: Feature Prompt Tuning for Few-shot Readability Assessment

83% match1 repo42024

Tailored-LLaMA: Optimizing Few-Shot Learning in Pruned LLaMA Models with Task-Specific Prompts

83% match2024

Automated Data Curation for Robust Language Model Fine-Tuning

82% match2024

Cite this paper

@article{najafi2024riff,
  title  = {RIFF: Learning to Rephrase Inputs for Few-shot Fine-tuning of Language Models},
  author = {Saeed Najafi and Alona Fyshe},
  year   = {2024},
  eprint = {2403.02271},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2403.02271v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2403.02271.svg)](https://codeofpaper.com/paper/2403.02271)

Discussion