Pseudo-OOD training for robust language models

TL;DR

This paper introduces POORE, a post hoc method that generates pseudo-OOD samples from in-distribution data to improve out-of-distribution detection in language models, achieving state-of-the-art results.

Contribution

The paper proposes a novel post hoc regularization framework that creates pseudo-OOD samples for better OOD detection without needing prior OOD data.

Findings

Significant improvements in OOD detection accuracy.

Achieved new state-of-the-art results on three dialogue systems.

Effective in real-world NLP applications.

Abstract

While pre-trained large-scale deep models have garnered attention as an important topic for many downstream natural language processing (NLP) tasks, such models often make unreliable predictions on out-of-distribution (OOD) inputs. As such, OOD detection is a key component of a reliable machine-learning model for any industry-scale application. Common approaches often assume access to additional OOD samples during the training stage, however, outlier distribution is often unknown in advance. Instead, we propose a post hoc framework called POORE - POsthoc pseudo-Ood REgularization, that generates pseudo-OOD samples using in-distribution (IND) data. The model is fine-tuned by introducing a new regularization loss that separates the embeddings of IND and OOD data, which leads to significant gains on the OOD prediction task during testing. We extensively evaluate our framework on three real-world dialogue systems, achieving new state-of-the-art in OOD detection.