ACCEPT: Adaptive Codebook for Composite and Efficient Prompt Tuning

TL;DR

ACCEPT introduces an adaptive codebook approach for prompt tuning that shares codebook vectors among prompts, significantly reducing parameters while maintaining high performance across diverse NLP tasks.

Contribution

The paper proposes ACCEPT, a novel prompt tuning method using shared codebooks and adaptive weights, enabling parameter-efficient tuning for large language models.

Findings

Achieves superior performance on 17 NLP tasks.

Tunes only 0.3% of parameters, demonstrating high efficiency.

Performs well in few-shot and large model scenarios.

Abstract

Prompt Tuning has been a popular Parameter-Efficient Fine-Tuning method attributed to its remarkable performance with few updated parameters on various large-scale pretrained Language Models (PLMs). Traditionally, each prompt has been considered indivisible and updated independently, leading the parameters increase proportionally as prompt length grows. To address this issue, we propose Adaptive Codebook for Composite and Efficient Prompt Tuning (ACCEPT). In our method, we refer to the concept of product quantization (PQ), allowing all soft prompts to share a set of learnable codebook vectors in each subspace, with each prompt differentiated by a set of adaptive weights. We achieve the superior performance on 17 diverse natural language tasks including natural language understanding (NLU) and question answering (QA) tasks by tuning only 0.3% of parameters of the PLMs. Our approach also excels in few-shot and large model settings, highlighting its significant potential.