Scaling up Discovery of Latent Concepts in Deep NLP Models
This work addresses the need for more efficient interpretability methods for deep NLP models, though it is incremental as it builds on existing clustering approaches.
This paper tackled the problem of scaling up the discovery of latent concepts in deep NLP models, which was previously limited by inefficient clustering methods, and found that using K-Means-based concept discovery significantly enhances efficiency while maintaining concept quality, enabling scaling to larger datasets and models like LLMs.
Despite the revolution caused by deep NLP models, they remain black boxes, necessitating research to understand their decision-making processes. A recent work by Dalvi et al. (2022) carried out representation analysis through the lens of clustering latent spaces within pre-trained models (PLMs), but that approach is limited to small scale due to the high cost of running Agglomerative hierarchical clustering. This paper studies clustering algorithms in order to scale the discovery of encoded concepts in PLM representations to larger datasets and models. We propose metrics for assessing the quality of discovered latent concepts and use them to compare the studied clustering algorithms. We found that K-Means-based concept discovery significantly enhances efficiency while maintaining the quality of the obtained concepts. Furthermore, we demonstrate the practicality of this newfound efficiency by scaling latent concept discovery to LLMs and phrasal concepts.