← Back to archive day

Parameter Importance is Not Static: Evolving Parameter Isolation for Supervised Fine-Tuning

Zekai Lin, Chao Xue, Di Liang, Xingsheng Han, Peiyang Liu, Xianjie Wu, Lei Jiang, Yu Lu, Haibo Shi, Shuang Liang, Minlong Peng

36

Recommendation Score

breakthrough🔴 AdvancedNLPFine-tuning & PEFTBenchmarkBest for researchers

Research context

Primary field

NLP

Language understanding, generation, extraction, and evaluation.

Topics

Fine-tuning & PEFT

Paper type

Benchmark

Best for

Best for researchers

arXiv categories

cs.LGcs.CLcs.LG

Why It Matters

Demonstrates parameter importance evolves during fine-tuning, introducing dynamic isolation that outperforms static PEFT methods—essential for efficient, stable LLM adaptation in production.

Abstract

Supervised Fine-Tuning (SFT) of large language models often suffers from task interference and catastrophic forgetting. Recent approaches alleviate this issue by isolating task-critical parameters during training. However, these methods represent a static solution to a dynamic problem, assuming that parameter importance remains fixed once identified. In this work, we empirically demonstrate that parameter importance exhibits temporal drift over the course of training. To address this, we propose Evolving Parameter Isolation (EPI), a fine-tuning framework that adapts isolation decisions based on online estimates of parameter importance. Instead of freezing a fixed subset of parameters, EPI periodically updates isolation masks using gradient-based signals, enabling the model to protect emerging task-critical parameters while releasing outdated ones to recover plasticity. Experiments on diverse multi-task benchmarks demonstrate that EPI consistently reduces interference and forgetting compared to static isolation and standard fine-tuning, while improving overall generalization. Our analysis highlights the necessity of synchronizing isolation mechanisms with the evolving dynamics of learning diverse abilities.

Published April 15, 2026
© 2026 A2A.pub — AI to Action. From papers to practice, daily.
Summaries are AI-assistedPrivacyTerms