iPOE: Interpretable Prompt Optimization via Explanations

TL;DR

iPOE introduces an interpretable prompt optimization method that uses explanations to generate guidelines, enhancing transparency and performance in large language model prompts across multiple datasets.

Contribution

The paper presents iPOE, a novel prompt optimization approach that incorporates explanations to create transparent, effective guidelines, improving prompt performance and interpretability.

Findings

iPOE improves prompt performance by up to 35% over baselines.

Guidelines generated from explanations can replace human explanations.

iPOE enhances transparency and supports laypeople in prompt optimization.

Abstract

Prompt optimization has often been framed as a discrete search problem to find high-performing and robust instructions for an LLM. However, the search result might not make it transparent why and where specific prompt changes lead to performance gains. This is in contrast to how humans are instructed for annotation tasks. Here, researchers carefully design annotation guidelines, leading to enhanced annotation consistency. Our paper aims at joining these two approaches and introduces iPOE, a novel interpretable prompt optimization strategy via explanations. We guide the prompt optimization process by automatically created guidelines from explanations of annotation decisions (either automatically generated or from humans). This set of guidelines is furthermore optimized by as series of operations, including removing, adding, shuffling, and merging. The resulting prompt includes guidelines that instruct the annotation, making the decision process of the LLM and the optimization transparent. It therefore supports also laypeople in the area of prompt optimization, particularly in challenging domains requiring expertise. In our experiments on four datasets, we find that iPOE can improves over prompts without guidelines and with random selected guidelines by up to $31\%$ and $35\%$, respectively. Moreover, LLM explanations can replace human explanations in the proposed method.