Publication Type

Conference Proceeding Article

Version

publishedVersion

Publication Date

12-2023

Abstract

Topic Modelling is an established research area where the quality of a given topic is measured using coherence metrics. Often, we infer topics from Neural Topic Models (NTM) by interpreting their decoder weights, consisting of top-activated words projected from individual neurons. Transformer-based Language Models (TLM) similarly consist of decoder weights. However, due to its hypothesised superposition properties, the final logits originating from the residual path are considered uninterpretable. Therefore, we posit that we can interpret TLM as superposed NTM by proposing a novel weight-based, model-agnostic and corpus-agnostic approach to search and disentangle decoder-only TLM, potentially mapping individual neurons to multiple coherent topics. Our results show that it is empirically feasible to disentangle coherent topics from GPT-2 models using the Wikipedia corpus. We validate this approach for GPT-2 models using Zero-Shot Topic Modelling. Finally, we extend the proposed approach to disentangle and analyse LLaMA models.

Discipline

Databases and Information Systems | Programming Languages and Compilers

Research Areas

Data Science and Engineering

Publication

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, Singapore, December 6-10

First Page

8646

Last Page

8666

Publisher

ACL

City or Country

Singapore

Copyright Owner and License

Authors

Additional URL

https://aclanthology.org/2023.emnlp-main.534

Share

COinS