Fixing Model Bugs with Natural Language Patches

TL;DR

This paper introduces a method for fixing NLP model bugs using natural language patches, which are declarative corrections that improve model accuracy with minimal data and effort.

Contribution

It proposes modeling patch applicability separately from integration, demonstrating effective use of natural language patches to fix models with few examples.

Findings

Natural language patches improve accuracy by 1-4 points on sentiment analysis.

F1 score increases by 7 points on relation extraction.

Small synthetic datasets suffice to teach models to use real patches effectively.

Abstract

Current approaches for fixing systematic problems in NLP models (e.g. regex patches, finetuning on more data) are either brittle, or labor-intensive and liable to shortcuts. In contrast, humans often provide corrections to each other through natural language. Taking inspiration from this, we explore natural language patches -- declarative statements that allow developers to provide corrective feedback at the right level of abstraction, either overriding the model (``if a review gives 2 stars, the sentiment is negative'') or providing additional information the model may lack (``if something is described as the bomb, then it is good''). We model the task of determining if a patch applies separately from the task of integrating patch information, and show that with a small amount of synthetic data, we can teach models to effectively use real patches on real data -- 1 to 7 patches improve accuracy by ~1-4 accuracy points on different slices of a sentiment analysis dataset, and F1 by 7 points on a relation extraction dataset. Finally, we show that finetuning on as many as 100 labeled examples may be needed to match the performance of a small set of language patches.