How Does Diffusion Influence Pretrained Language Models on Out-of-Distribution Data?

TL;DR

This paper investigates how applying diffusion processes to pretrained language models affects their robustness to out-of-distribution data, revealing that diffusion can both impair reconstruction ability and enhance OOD detection.

Contribution

It provides the first comprehensive analysis of diffusion's impact on PLMs' OOD robustness, including evaluation of reconstruction and detection capabilities.

Findings

Diffusion training degrades OOD reconstruction ability.

Diffusion models improve OOD sample detection, achieving state-of-the-art accuracy.

Diffusion reduces overall OOD robustness of PLMs.

Abstract

Transformer-based pretrained language models (PLMs) have achieved great success in modern NLP. An important advantage of PLMs is good out-of-distribution (OOD) robustness. Recently, diffusion models have attracted a lot of work to apply diffusion to PLMs. It remains under-explored how diffusion influences PLMs on OOD data. The core of diffusion models is a forward diffusion process which gradually applies Gaussian noise to inputs, and a reverse denoising process which removes noise. The noised input reconstruction is a fundamental ability of diffusion models. We directly analyze OOD robustness by measuring the reconstruction loss, including testing the abilities to reconstruct OOD data, and to detect OOD samples. Experiments are conducted by analyzing different training parameters and data statistical features on eight datasets. It shows that finetuning PLMs with diffusion degrades the reconstruction ability on OOD data. The comparison also shows that diffusion models can effectively detect OOD samples, achieving state-of-the-art performance in most of the datasets with an absolute accuracy improvement up to 18%. These results indicate that diffusion reduces OOD robustness of PLMs.