Faithfulness Measurable Masked Language Models

arXiv:2310.07819v2 Announce Type: replace-cross
Abstract: A common approach to explaining NLP models is to use importance measures that express which tokens are important for a prediction. Unfortunately, such explanations are often wrong despite being persuasive. Therefore, it is essential to measure their faithfulness. One such metric is if tokens are truly important, then masking them should result in worse model performance. However, token masking introduces out-of-distribution issues, and existing solutions that address this are computationally expensive and employ proxy models. …

abstract arxiv cs.cl cs.lg express importance language language models masking nlp nlp models prediction replace them tokens type