Information-Theoretic Generative Clustering of Documents
This addresses document clustering and retrieval for users needing improved accuracy, though it appears incremental as it builds on existing LLM capabilities.
The paper tackles document clustering by using texts generated by large language models (LLMs) to define similarity via KL divergence, achieving state-of-the-art performance with large margins over previous methods.
We present {\em generative clustering} (GC) for clustering a set of documents, $\mathrm{X}$, by using texts $\mathrm{Y}$ generated by large language models (LLMs) instead of by clustering the original documents $\mathrm{X}$. Because LLMs provide probability distributions, the similarity between two documents can be rigorously defined in an information-theoretic manner by the KL divergence. We also propose a natural, novel clustering algorithm by using importance sampling. We show that GC achieves the state-of-the-art performance, outperforming any previous clustering method often by a large margin. Furthermore, we show an application to generative document retrieval in which documents are indexed via hierarchical clustering and our method improves the retrieval accuracy.