Logit-Based Ensemble Distribution Distillation for Robust Autoregressive Sequence Uncertainties

TL;DR

This paper introduces a logit-based ensemble distribution distillation method for autoregressive sequence models, improving uncertainty estimation and out-of-distribution detection while reducing computational costs.

Contribution

It proposes modeling ensemble logits instead of probabilities, enabling scalable uncertainty distillation for large vocabularies in transformer models.

Findings

Logit-based EDD outperforms probability-space methods.

Students surpass Deep Ensembles in OOD detection by ~10%.

Method maintains in-distribution translation quality.

Abstract

Efficiently and reliably estimating uncertainty is an important objective in deep learning. It is especially pertinent to autoregressive sequence tasks, where training and inference costs are typically very high. However, existing research has predominantly focused on tasks with static data such as image classification. In this work, we investigate Ensemble Distribution Distillation (EDD) applied to large-scale natural language sequence-to-sequence data. EDD aims to compress the superior uncertainty performance of an expensive (teacher) ensemble into a cheaper (student) single model. Importantly, the ability to separate knowledge (epistemic) and data (aleatoric) uncertainty is retained. Existing probability-space approaches to EDD, however, are difficult to scale to large vocabularies. We show, for modern transformer architectures on large-scale translation tasks, that modelling the ensemble logits, instead of softmax probabilities, leads to significantly better students. Moreover, the students surprisingly even outperform Deep Ensembles by up to ~10% AUROC on out-of-distribution detection, whilst matching them at in-distribution translation.