Mitigate Negative Transfer with Similarity Heuristic Lifelong Prompt Tuning

TL;DR

This paper introduces SHLPT, a framework that improves lifelong prompt tuning by using a learnable similarity metric to reduce negative transfer and enhance task transferability.

Contribution

The paper proposes a novel similarity heuristic approach that partitions tasks and employs a parameter pool to mitigate negative transfer and catastrophic forgetting in lifelong prompt tuning.

Findings

SHLPT outperforms state-of-the-art methods in benchmarks.

SHLPT effectively reduces negative transfer across diverse tasks.

The similarity heuristic improves task transferability regardless of task similarity.

Abstract

Lifelong prompt tuning has significantly advanced parameter-efficient lifelong learning with its efficiency and minimal storage demands on various tasks. Our empirical studies, however, highlights certain transferability constraints in the current methodologies: a universal algorithm that guarantees consistent positive transfer across all tasks is currently unattainable, especially when dealing dissimilar tasks that may engender negative transfer. Identifying the misalignment between algorithm selection and task specificity as the primary cause of negative transfer, we present the Similarity Heuristic Lifelong Prompt Tuning (SHLPT) framework. This innovative strategy partitions tasks into two distinct subsets by harnessing a learnable similarity metric, thereby facilitating fruitful transfer from tasks regardless of their similarity or dissimilarity. Additionally, SHLPT incorporates a parameter pool to combat catastrophic forgetting effectively. Our experiments shows that SHLPT outperforms state-of-the-art techniques in lifelong learning benchmarks and demonstrates robustness against negative transfer in diverse task sequences.