Investigating Zero- and Few-shot Generalization in Fact Verification
This work addresses the challenge of applying fact verification models to domains with limited annotations, which is incremental as it builds on existing methods with new empirical insights and benchmark creation.
The paper tackled the problem of poor generalization in fact verification models from well-resourced to low-resourced domains by constructing a benchmark of 11 datasets across 6 domains and identifying factors like dataset size and evidence length that affect performance. It showed that incorporating domain knowledge through pretraining and generating training data via claim generation can improve generalization, though specific numerical gains were not provided.
In this paper, we explore zero- and few-shot generalization for fact verification (FV), which aims to generalize the FV model trained on well-resourced domains (e.g., Wikipedia) to low-resourced domains that lack human annotations. To this end, we first construct a benchmark dataset collection which contains 11 FV datasets representing 6 domains. We conduct an empirical analysis of generalization across these FV datasets, finding that current models generalize poorly. Our analysis reveals that several factors affect generalization, including dataset size, length of evidence, and the type of claims. Finally, we show that two directions of work improve generalization: 1) incorporating domain knowledge via pretraining on specialized domains, and 2) automatically generating training data via claim generation.