Can LLMs facilitate interpretation of pre-trained language models? (2305.13386v2)

Abstract: Work done to uncover the knowledge encoded within pre-trained LLMs rely on annotated corpora or human-in-the-loop methods. However, these approaches are limited in terms of scalability and the scope of interpretation. We propose using a LLM, ChatGPT, as an annotator to enable fine-grained interpretation analysis of pre-trained LLMs. We discover latent concepts within pre-trained LLMs by applying agglomerative hierarchical clustering over contextualized representations and then annotate these concepts using ChatGPT. Our findings demonstrate that ChatGPT produces accurate and semantically richer annotations compared to human-annotated concepts. Additionally, we showcase how GPT-based annotations empower interpretation analysis methodologies of which we demonstrate two: probing frameworks and neuron interpretation. To facilitate further exploration and experimentation in the field, we make available a substantial ConceptNet dataset (TCN) comprising 39,000 annotated concepts.