Multiple Queries with Multiple Keys: A Precise Prompt Matching Paradigm for Prompt-based Continual Learning

TL;DR

This paper introduces MQMK, a prompt matching paradigm that improves prompt selection accuracy in continual learning by using multiple queries and keys to closely match training and test data distributions, leading to state-of-the-art results.

Contribution

The paper proposes a novel MQMK paradigm that enhances prompt matching precision in continual learning through a dual-query and key mechanism, addressing bias issues in prompt selection.

Findings

Increases prompt matching rate by over 30% in challenging scenarios.

Achieves state-of-the-art performance on three continual learning benchmarks.

Demonstrates effective reduction of bias in prompt-based continual learning.

Abstract

Continual learning requires machine learning models to continuously acquire new knowledge in dynamic environments while avoiding the forgetting of previous knowledge. Prompt-based continual learning methods effectively address the issue of catastrophic forgetting through prompt expansion and selection. However, existing approaches often suffer from low accuracy in prompt selection, which can result in the model receiving biased knowledge and making biased predictions. To address this issue, we propose the Multiple Queries with Multiple Keys (MQMK) prompt matching paradigm for precise prompt selection. The goal of MQMK is to select the prompts whose training data distribution most closely matches that of the test sample. Specifically, Multiple Queries enable precise breadth search by introducing task-specific knowledge, while Multiple Keys perform deep search by representing the feature distribution of training samples at a fine-grained level. Each query is designed to perform local matching with a designated task to reduce interference across queries. Experiments show that MQMK enhances the prompt matching rate by over 30\% in challenging scenarios and achieves state-of-the-art performance on three widely adopted continual learning benchmarks. The code is available at https://github.com/DunweiTu/MQMK.