Cluster-R1: Large Reasoning Models Are Instruction-following Clustering Agents

TL;DR

This paper introduces a new approach called Cluster-R1, where large reasoning models are trained as autonomous clustering agents capable of following instructions and inferring latent data structures, outperforming traditional embedding methods.

Contribution

The paper presents a novel training pipeline for large reasoning models to interpret instructions and infer clustering structures, bridging the gap between embedding models and autonomous clustering agents.

Findings

LRMs outperform embedding-based methods in clustering tasks.

Explicit reasoning improves interpretability and faithfulness of clustering.

The ReasonCluster benchmark evaluates diverse real-world clustering scenarios.

Abstract

General-purpose embedding models excel at recognizing semantic similarities but fail to capture the characteristics of texts specified by user instructions. In contrast, instruction-tuned embedders can align embeddings with textual instructions yet cannot autonomously infer latent corpus structures, such as determining the optimal number of clusters. To address both limitations, we reframe instruction-following clustering as a generative task and train large reasoning models (LRMs) as autonomous clustering agents. Our reasoning-driven training pipeline enables LRMs to interpret high-level clustering instructions and then infer the corresponding latent groupings. To evaluate this paradigm, we introduce ReasonCluster, a comprehensive benchmark comprising 28 diverse tasks spanning daily dialogue, legal cases, and financial reports. Experiments across diverse datasets and clustering scenarios show that our approach consistently outperforms strong embedding-based methods and LRM baselines, demonstrating that explicit reasoning fosters more faithful and interpretable instruction-based clustering.