Logit-Based Ensemble Distribution Distillation for Robust Autoregressive Sequence Uncertainties
This work addresses uncertainty estimation for autoregressive sequence tasks like translation, which is incremental as it adapts an existing method to a new domain with specific improvements.
The paper tackled the problem of efficiently estimating uncertainty in autoregressive sequence tasks by applying Ensemble Distribution Distillation to large-scale natural language sequence-to-sequence data, resulting in students that outperform Deep Ensembles by up to ~10% AUROC on out-of-distribution detection while matching them at in-distribution translation.
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.