FactPEGASUS: Factuality-Aware Pre-training and Fine-tuning for Abstractive Summarization
This addresses the critical issue of generating factual summaries for applications like news and research, though it is an incremental improvement over existing models like PEGASUS.
The authors tackled the problem of factuality in abstractive summarization by developing FactPEGASUS, which modifies pre-training and fine-tuning to reduce hallucinations, resulting in substantial improvements in factuality on three downstream tasks as evaluated by multiple metrics and humans.
We present FactPEGASUS, an abstractive summarization model that addresses the problem of factuality during pre-training and fine-tuning: (1) We augment the sentence selection strategy of PEGASUS's (Zhang et al., 2020) pre-training objective to create pseudo-summaries that are both important and factual; (2) We introduce three complementary components for fine-tuning. The corrector removes hallucinations present in the reference summary, the contrastor uses contrastive learning to better differentiate nonfactual summaries from factual ones, and the connector bridges the gap between the pre-training and fine-tuning for better transfer of knowledge. Experiments on three downstream tasks demonstrate that FactPEGASUS substantially improves factuality evaluated by multiple automatic metrics and humans. Our thorough analysis suggests that FactPEGASUS is more factual than using the original pre-training objective in zero-shot and few-shot settings, retains factual behavior more robustly than strong baselines, and does not rely entirely on becoming more extractive to improve factuality. Our code and data are publicly available at: https://github.com/meetdavidwan/factpegasus