CATfOOD: Counterfactual Augmented Training for Improving Out-of-Domain Performance and Calibration

TL;DR

This paper introduces a method using large language models to generate counterfactual data for small models, significantly enhancing out-of-domain question answering performance and calibration by increasing data diversity and explanation conciseness.

Contribution

It presents a novel data augmentation technique with counterfactual instances generated by LLMs to improve out-of-domain performance and calibration of small language models.

Findings

Counterfactual augmentation improves OOD QA performance.

Enhanced calibration correlates with diverse CF instances.

Calibrated models show lower entropy in importance attribution.

Abstract

In recent years, large language models (LLMs) have shown remarkable capabilities at scale, particularly at generating text conditioned on a prompt. In our work, we investigate the use of LLMs to augment training data of small language models~(SLMs) with automatically generated counterfactual~(CF) instances -- i.e. minimally altered inputs -- in order to improve out-of-domain~(OOD) performance of SLMs in the extractive question answering~(QA) setup. We show that, across various LLM generators, such data augmentation consistently enhances OOD performance and improves model calibration for both confidence-based and rationale-augmented calibrator models. Furthermore, these performance improvements correlate with higher diversity of CF instances in terms of their surface form and semantic content. Finally, we show that CF augmented models which are easier to calibrate also exhibit much lower entropy when assigning importance, indicating that rationale-augmented calibrators prefer concise explanations.