Agentic Clustering: Controllable Text Taxonomies via Multi-Agent Refinement
It addresses the problem of rigid, programmatic clustering pipelines that fail to generalize across corpora and cannot incorporate user constraints, offering a more adaptable and controllable approach for text clustering.
The paper introduces an agentic text-clustering method that uses an orchestrator LLM to dynamically dispatch specialized agents, achieving state-of-the-art performance on seven benchmarks with up to 32% improvement in ARI over prior LLM baselines.
Recent text-clustering methods use large language models to propose a cluster taxonomy from a corpus and then assign each text to it. These pipelines are fundamentally programmatic: the sequence of LLM calls and the rules for stopping, merging, and splitting clusters are fixed in code in advance, so they generalise poorly across corpora of different structure and cannot easily incorporate user-supplied constraints such as a target cluster count or a clustering intent. We propose an agentic alternative in which an orchestrator LLM inspects the state of the discovery process at each step and dispatches one of a small set of specialised agents - proposer, synthesizer, auditor, investigator, and critic - adapting the pipeline to the corpus rather than executing a fixed one. On seven public text-clustering benchmarks the method achieves state-of-the-art performance, beating the strongest prior LLM baseline by up to 32% in ARI.