Semantic are Beacons: A Semantic Perspective for Unveiling Parameter-Efficient Fine-Tuning in Knowledge Learning

TL;DR

This paper investigates why parameter-efficient fine-tuning struggles with factual knowledge learning in large language models, revealing semantic interference issues and proposing strategies to improve knowledge retention.

Contribution

It introduces a semantic perspective to analyze PEFT limitations and proposes data filtering and re-weighted learning strategies to enhance knowledge learning in LLMs.

Findings

PEFT can push models away from knowledge targets

Multiple knowledge sources interfere, hindering learning

Proposed strategies improve knowledge retention in experiments

Abstract

Parameter-Efficient Fine-Tuning (PEFT) methods enable efficient adaptation of Large Language Models (LLMs) to various downstream applications. However, the effectiveness of the PEFT diminishes notably when downstream tasks require accurate learning of factual knowledge. In this paper, we adopt a semantic perspective to investigate this phenomenon, uncovering the reasons behind PEFT's limitations in knowledge learning task. Our findings reveal that: (1) PEFT presents a notable risk of pushing the model away from the intended knowledge target; (2) multiple knowledge interfere with each other, and such interference suppresses the learning and expression of knowledge features. Based on these insights, we introduce a data filtering strategy to exclude data that is detrimental to knowledge learning and a re-weighted learning strategy to make the model attentive to semantic distance during knowledge learning. Experimental results demonstrate the effectiveness of the proposed method on open-source large language model, further validate the semantic challenge in PEFT, thus paving the way for future research.