EMNLP 2025 paper: Efficient Layer-wise LLM Fine-tuning for Revision Intention Prediction
Abstract: Large Language Models (LLMs) have shown extraordinary success across various text generation tasks; however, their potential for simple yet essential text classification remains underexplored, as LLM pre-training tends to emphasize generation over classification. While LLMs with instruction tuning can transform classification into a generation task, they often struggle to categorize nuanced texts. One such example is text revision, which involves nuanced edits between pairs of texts. Although simply fine-tuning LLMs for revision classification seems plausible, it requires a large amount of revision annotations, which are exceptionally expensive and scarce in the community. To address this issue, we introduce a plug-and-play layer-wise parameter-efficient fine-tuning (PEFT) framework, i.e., IR-Tuning, which fine-tunes a subset of important LLM layers that are dynamically selected based on their gradient norm distribution, while freezing those of redundant layers. Extensive experiments suggest that IR-Tuning surpasses several layer-wise PEFT baselines over diverse text revisions, while achieving fast convergence, low GPU memory consumption, and effectiveness on small revision corpora.
Please install Anaconda 24.5.0 with Python 3.9 and create a new virtual environment named revision using the yaml file and command conda env create -f environment.yaml -n revision. Please configure huggingface key and wandb key in config.yaml if needed and then run the following command to start fine-tuning. Note that you can change the hyperparameters in the command and more in the main.py file as needed.
python main.py \
--model-name llama3.1-8b \
--dataset-name iterater-human \
--adapter-name lora \
--importance-metric-name gradient \
--tuning-method ir \
--batch-size 8 \
--per-device-train-batch-size 8 \
--use-instruction
@inproceedings{
liu-litman-2025-efficient,
title = "Efficient Layer-wise {LLM} Fine-tuning for Revision Intention Prediction",
author = "Liu, Zhexiong and Litman, Diane",
editor = "Christodoulopoulos, Christos and Chakraborty, Tanmoy and Rose, Carolyn and Peng, Violet",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
month = Nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-emnlp.829/",
doi = "10.18653/v1/2025.findings-emnlp.829",
pages = "15319--15334",
ISBN = "979-8-89176-335-7"}
This project is built on prior IST codebase.