매일 아침, 어제의 AI를 한 통으로 정리해 보내드립니다메일로 받아보기

METAL LAB

ReCBM: Uncertainty-Gated Relational Reasoning for Concept Bottleneck Models

arXiv:2608.100042026-08-12

arXiv:2608.10004v1 Announce Type: new Abstract: Concept Bottleneck Models (CBMs) provide an interpretable framework by grounding predictions in human-understandable concepts, enabling semantic inspection and test-time intervention. Recent variants have improved CBMs through richer concept representations, uncertainty estimation, and dependency modeling. However, robust reasoning under unreliable concept states remains underexplored. Without such reasoning, misleading semantic evidence can propag

저자 · An Sui, Yuzhu Li, Fuping Wu, Xiahai Zhuang

arXiv에서 원문 보기