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

Researchers have developed Cluster-R1, a new approach that trains large reasoning models (LRMs) as autonomous clustering agents capable of following instructions and inferring optimal cluster structures. The method reframes instruction-following clustering as a generative task and demonstrates superior performance over traditional embedding-based methods across 28 diverse tasks in the ReasonCluster benchmark.

• →Large reasoning models can be trained as autonomous clustering agents that interpret high-level instructions and infer latent data groupings.
• →The approach addresses limitations of both general-purpose embedding models and instruction-tuned embedders by combining reasoning capabilities.
• →ReasonCluster benchmark includes 28 diverse tasks spanning dialogue, legal cases, and financial reports for comprehensive evaluation.
• →The reasoning-driven method consistently outperforms existing embedding-based clustering approaches across various scenarios.
• →Explicit reasoning enables more faithful and interpretable instruction-based clustering compared to traditional methods.