Out-of-Distribution Detection and Selective Generation for Conditional Language Models
This addresses the issue of unreliable and overly confident outputs in generative language models, particularly under distribution shift, enabling safer deployment for users in applications like summarization and translation.
The paper tackles the problem of out-of-distribution (OOD) inputs causing degradation in conditional language models, presenting a highly accurate and lightweight OOD detection method that enables selective generation of high-quality outputs while abstaining from low-quality ones, demonstrated on abstractive summarization and translation tasks.
Machine learning algorithms typically assume independent and identically distributed samples in training and at test time. Much work has shown that high-performing ML classifiers can degrade significantly and provide overly-confident, wrong classification predictions, particularly for out-of-distribution (OOD) inputs. Conditional language models (CLMs) are predominantly trained to classify the next token in an output sequence, and may suffer even worse degradation on OOD inputs as the prediction is done auto-regressively over many steps. Furthermore, the space of potential low-quality outputs is larger as arbitrary text can be generated and it is important to know when to trust the generated output. We present a highly accurate and lightweight OOD detection method for CLMs, and demonstrate its effectiveness on abstractive summarization and translation. We also show how our method can be used under the common and realistic setting of distribution shift for selective generation (analogous to selective prediction for classification) of high-quality outputs, while automatically abstaining from low-quality ones, enabling safer deployment of generative language models.