Interpreting Key Mechanisms of Factual Recall in Transformer-Based Language Models

arXiv:2403.19521v1 Announce Type: cross
Abstract: In this paper, we deeply explore the mechanisms employed by Transformer-based language models in factual recall tasks. In zero-shot scenarios, given a prompt like "The capital of France is," task-specific attention heads extract the topic entity, such as "France," from the context and pass it to subsequent MLPs to recall the required answer such as "Paris." We introduce a novel analysis method aimed at decomposing the outputs of the MLP into components understandable by humans. …

abstract arxiv attention capital context cs.ai cs.cl cs.lg explore extract france key language language models paper prompt recall tasks transformer type zero-shot