D.Va: Validate Your Demonstration First Before You Use It

TL;DR

The paper introduces D.Va, a demonstration validation method for in-context learning that improves demonstration selection robustness and generalization across tasks and models.

Contribution

D.Va is a novel demonstration validation approach that enhances demonstration selection for in-context learning, outperforming existing methods in robustness and cross-model generalization.

Findings

D.Va surpasses existing demonstration selection techniques in NLU and NLG tasks.

D.Va demonstrates robustness across various language models.

The method improves generalization in demonstration effectiveness.

Abstract

In-context learning (ICL) has demonstrated significant potential in enhancing the capabilities of large language models (LLMs) during inference. It's well-established that ICL heavily relies on selecting effective demonstrations to generate outputs that better align with the expected results. As for demonstration selection, previous approaches have typically relied on intuitive metrics to evaluate the effectiveness of demonstrations, which often results in limited robustness and poor cross-model generalization capabilities. To tackle these challenges, we propose a novel method, \textbf{D}emonstration \textbf{VA}lidation (\textbf{D.Va}), which integrates a demonstration validation perspective into this field. By introducing the demonstration validation mechanism, our method effectively identifies demonstrations that are both effective and highly generalizable. \textbf{D.Va} surpasses all existing demonstration selection techniques across both natural language understanding (NLU) and natural language generation (NLG) tasks. Additionally, we demonstrate the robustness and generalizability of our approach across various language models with different retrieval models.