Fairness-guided Few-shot Prompting for Large Language Models

Changqing Zhang, Yatao Bian, Lemao Liu, Zhirui Zhang, Peilin Zhao, Shu Zhang, Huazhu Fu, Qinghua Hu, Bingzhe Wu

arXiv (Cornell University) · 2023 · 24 citations · 0 references

DOIFull text

Open access

Concepts

TL;DR

Large language models can perform in‑context learning via prompts, but their performance is unstable due to variations in training examples, order, and format, making prompt construction essential. This work revisits prompt design from a predictive‑bias perspective and proposes a greedy‑search strategy to identify near‑optimal prompts for in‑context learning. We introduce a metric that quantifies a prompt’s predictive bias, use it to guide a greedy search for near‑optimal prompts, and evaluate the approach on GPT‑3 across multiple downstream tasks. Prompts with higher bias consistently yield poorer predictive quality, while our bias‑guided greedy search improves GPT‑3’s in‑context learning performance in an effective and interpretable manner.

Abstract

Large language models have demonstrated surprising ability to perform in-context learning, i.e., these models can be directly applied to solve numerous downstream tasks by conditioning on a prompt constructed by a few input-output examples. However, prior research has shown that in-context learning can suffer from high instability due to variations in training examples, example order, and prompt formats. Therefore, the construction of an appropriate prompt is essential for improving the performance of in-context learning. In this paper, we revisit this problem from the view of predictive bias. Specifically, we introduce a metric to evaluate the predictive bias of a fixed prompt against labels or a given attributes. Then we empirically show that prompts with higher bias always lead to unsatisfactory predictive quality. Based on this observation, we propose a novel search strategy based on the greedy search to identify the near-optimal prompt for improving the performance of in-context learning. We perform comprehensive experiments with state-of-the-art mainstream models such as GPT-3 on various downstream tasks. Our results indicate that our method can enhance the model's in-context learning performance in an effective and interpretable manner.