StablePrompt: Automatic Prompt Tuning using Reinforcement Learning for Large Language Models

TL;DR

StablePrompt introduces an RL-based prompt tuning method with adaptive updates to improve stability and performance across multiple NLP tasks in large language models.

Contribution

It proposes StablePrompt, a novel RL framework with APPO and an LLM anchor to enhance prompt tuning stability and effectiveness.

Findings

Outperforms previous prompt tuning methods on multiple NLP tasks

Achieves higher stability in reinforcement learning-based prompt search

Maintains linguistic capabilities of pre-trained LLMs

Abstract

Finding appropriate prompts for the specific task has become an important issue as the usage of Large Language Models (LLM) has expanded. Reinforcement Learning (RL) is widely used for prompt tuning, but its inherent instability and environmental dependency make it difficult to use in practice. In this paper, we propose StablePrompt, which strikes a balance between training stability and search space, mitigating the instability of RL and producing high-performance prompts. We formulate prompt tuning as an online RL problem between the agent and target LLM and introduce Adaptive Proximal Policy Optimization (APPO). APPO introduces an LLM anchor model to adaptively adjust the rate of policy updates. This allows for flexible prompt search while preserving the linguistic ability of the pre-trained LLM. StablePrompt outperforms previous methods on various tasks including text classification, question answering, and text generation. Our code can be found in github.