GAPX: Generalized Autoregressive Paraphrase-Identification X

TL;DR

GAPX introduces a novel approach to paraphrase identification by training separate models for positive and negative pairs, using a perplexity-based metric to balance their influence and improve robustness against distribution shifts.

Contribution

The paper proposes training separate models for positive and negative pairs and introduces a perplexity-based metric to optimally combine them during inference.

Findings

Effective reduction of performance drop under distribution shift

Perplexity-based metric accurately determines model weighting

Empirical results demonstrate improved robustness

Abstract

Paraphrase Identification is a fundamental task in Natural Language Processing. While much progress has been made in the field, the performance of many state-of-the-art models often suffer from distribution shift during inference time. We verify that a major source of this performance drop comes from biases introduced by negative examples. To overcome these biases, we propose in this paper to train two separate models, one that only utilizes the positive pairs and the other the negative pairs. This enables us the option of deciding how much to utilize the negative model, for which we introduce a perplexity based out-of-distribution metric that we show can effectively and automatically determine how much weight it should be given during inference. We support our findings with strong empirical results.