매일 아침, 어제의 AI를 한 통으로 정리해 보내드립니다메일로 받아보기

METAL LAB

GRACE: LLM-Grounded Semantic Metric Spaces for Scalable Mixed-Data Clustering

arXiv:2608.078812026-08-11

arXiv:2608.07881v1 Announce Type: new Abstract: Clustering mixed tabular data requires a unified metric space to bridge the inherent heterogeneity between continuous numerical measurements and discrete categorical symbols. Traditionally, algorithms rely entirely on dataset-internal statistics to estimate categorical relationships, which confines the learned metric to empirical co-occurrences and ignores conceptually obvious yet statistically unobserved affinities. Although LLMs offer external wo

저자 · Zihua Yang, Zhencheng Xie, Junyang Chen, Liang Xie, Yiqun Zhang, Mengke Li, Yang Lu

arXiv에서 원문 보기