On the Utility of Domain-Adjacent Fine-Tuned Model Ensembles for Few-shot Problems

TL;DR

This paper introduces DAFT-E, an ensemble framework of domain-adjacent fine-tuned models that enhances few-shot learning performance, approaching or surpassing single models with less data.

Contribution

The paper proposes DAFT-E, a novel ensemble method that leverages domain-adjacent models to improve few-shot and zero-shot task performance.

Findings

Ensembling domain-adjacent models achieves near-optimal zero-shot accuracy.

Few-shot performance improves significantly with DAFT-E, surpassing individual models.

DAFT-E requires less domain-specific data than traditional fine-tuning methods.

Abstract

Large Language Models (LLMs) have been observed to perform well on a wide range of downstream tasks when fine-tuned on domain-specific data. However, such data may not be readily available in many applications, motivating zero-shot or few-shot approaches using domain-adjacent models. While several fine-tuned models for various tasks are available, finding an appropriate domain-adjacent model for a given task is often not straight forward. In this paper, we study DAFT-E, a framework that utilizes an Ensemble of Domain-Adjacent Fine-Tuned Foundation Models for few-shot problems. We show that for zero-shot problems, this ensembling method provides an accuracy performance close to that of the single best model. With few-shot problems, this performance improves further, at which point DEFT-E can outperform any single domain-adjacent model while requiring much less data for domain-specific fine-tuning.