Probing of Quantitative Values in Abstractive Summarization Models

TL;DR

This paper investigates how well current abstractive summarization models understand and represent quantitative data, revealing that most models struggle with accurate encoding, which may contribute to data hallucination issues.

Contribution

It introduces probing tests to evaluate quantitative value representation in summarization models and analyzes the impact of encoder performance on hallucination problems.

Findings

Most models' encoders poorly represent quantitative values.

DistilBART-CDM underperforms compared to random baselines.

Pretraining and fine-tuning may influence encoder effectiveness.

Abstract

Abstractive text summarization has recently become a popular approach, but data hallucination remains a serious problem, including with quantitative data. We propose a set of probing tests to evaluate the efficacy of abstract summarization models' modeling of quantitative values found in the input text. Our results show that in most cases, the encoders of recent SOTA-performing models struggle to provide embeddings that adequately represent quantitative values in the input compared to baselines, and in particular, they outperform random representations in some, but surprisingly not all, cases. Under our assumptions, this suggests that the encoder's performance contributes to the quantity hallucination problem. One model type in particular, DistilBART-CDM, was observed to underperform randomly initialized representations for several experiments, and performance versus BERT suggests that standard pretraining and fine-tuning approaches for the summarization task may play a role in underperformance for some encoders.