Yesterday's News: Benchmarking Multi-Dimensional Out-of-Distribution Generalization of Misinformation Detection Models

TL;DR

This paper presents misinfo-general, a benchmark dataset designed to evaluate misinformation detection models' ability to generalize across different out-of-distribution scenarios, addressing the rapid evolution of misinformation content.

Contribution

It introduces a novel benchmark dataset with simulated covariate shifts for assessing out-of-distribution generalization in misinformation detection models.

Findings

Baseline models perform poorly on out-of-distribution shifts.

Metadata can reveal model failures not evident from classification metrics.

The dataset minimizes shortcuts, ensuring robust evaluation.

Abstract

This article introduces misinfo-general, a benchmark dataset for evaluating misinformation models' ability to perform out-of-distribution generalization. Misinformation changes rapidly, much more quickly than moderators can annotate at scale, resulting in a shift between the training and inference data distributions. As a result, misinformation detectors need to be able to perform out-of-distribution generalization, an attribute they currently lack. Our benchmark uses distant labelling to enable simulating covariate shifts in misinformation content. We identify time, event, topic, publisher, political bias, misinformation type as important axes for generalization, and we evaluate a common class of baseline models on each. Using article metadata, we show how this model fails desiderata, which is not necessarily obvious from classification metrics. Finally, we analyze properties of the data to ensure limited presence of modelling shortcuts. We make the dataset and accompanying code publicly available: https://github.com/ioverho/misinfo-general