ReCBM: Uncertainty-Gated Relational Reasoning for Concept Bottleneck Models

ReCBM: Uncertainty-Gated Relational Reasoning for Concept Bottleneck Models

ReCBM：面向概念瓶颈模型的不确定性门控关系推理

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 propagate through the bottleneck, compromising both explanations and downstream predictions.

摘要： 概念瓶颈模型（CBMs）通过将预测建立在人类可理解的概念之上，提供了一种可解释的框架，从而实现了语义检查和测试时的干预。近期的变体通过更丰富的概念表示、不确定性估计和依赖建模改进了 CBM。然而，在不可靠的概念状态下进行稳健推理的研究仍显不足。缺乏这种推理机制时，误导性的语义证据可能会在瓶颈中传播，从而损害解释的准确性和下游预测的性能。

To address this issue, we propose ReCBM, an uncertainty-gated relational reasoning framework for CBMs. ReCBM introduces semantically defined concept relations into the bottleneck and uses uncertainty to guide their refinement. By modeling co-occurrence, implication, and exclusion, ReCBM specifies how evidence is exchanged across concepts, while uncertainty modulates the contribution of each concept during this process.

为了解决这一问题，我们提出了 ReCBM，这是一个面向 CBM 的不确定性门控关系推理框架。ReCBM 将语义定义的概念关系引入瓶颈层，并利用不确定性来指导其优化。通过对共现、蕴含和互斥关系进行建模，ReCBM 明确了证据如何在不同概念间进行交换，同时在这一过程中，不确定性调节了每个概念的贡献度。

Experiments across diverse datasets showed that ReCBM improved concept and task recovery under missing and flipped concepts, supported uncertainty-aware intervention, and extracted compact task-relevant concept subsets without degrading downstream performance.

在多个数据集上的实验表明，ReCBM 在概念缺失或翻转的情况下改善了概念和任务的恢复效果，支持了基于不确定性的干预，并能够在不降低下游性能的前提下，提取出紧凑且与任务相关的概念子集。