Few-Shot Self-Rationalization with Natural Language Prompts
This addresses the challenge of making self-rationalization models more accessible for NLP systems by reducing data requirements, though it is incremental as it builds on existing datasets and methods.
The paper tackles the problem of training self-rationalization models, which predict labels and generate explanations, by proposing a few-shot setting to reduce reliance on large human-written datasets, and shows that using natural language prompts and scaling model size can achieve progress, with GPT-3 reaching 51% plausibility in explanations compared to 76% for humans.
Self-rationalization models that predict task labels and generate free-text elaborations for their predictions could enable more intuitive interaction with NLP systems. These models are, however, currently trained with a large amount of human-written free-text explanations for each task which hinders their broader usage. We propose to study a more realistic setting of self-rationalization using few training examples. We present FEB -- a standardized collection of four existing English-language datasets and associated metrics. We identify the right prompting approach by extensively exploring natural language prompts on FEB. Then, by using this prompt and scaling the model size, we demonstrate that making progress on few-shot self-rationalization is possible. We show there is still ample room for improvement in this task: the average plausibility of generated explanations assessed by human annotators is at most 51% (with GPT-3), while plausibility of human explanations is 76%. We hope that FEB and our proposed approach will spur the community to take on the few-shot self-rationalization challenge.