The paper "BEARS Make Neuro-Symbolic Models Aware of Their Reasoning Shortcuts" by Emanuele Marconato et al. addresses the issue of Reasoning Shortcuts (RSs) in Neuro-Symbolic (NeSy) models, which can lead to overconfidence and unreliable predictions. RSs occur when models learn concepts with unintended semantics, compromising reliability and generalization. The authors propose a method called BEARS (BE Aware of Reasoning Shortcuts) to make NeSy models aware of RSs by calibrating their concept-level confidence without sacrificing prediction accuracy. BEARS achieves this by ensembleizing concept extractors that are diverse in their predictions of concepts affected by RSs. Empirical results show that BEARS improves RS-awareness in state-of-the-art NeSy models and facilitates the acquisition of informative dense annotations for mitigation purposes. The paper also discusses the benefits and limitations of BEARS, highlighting its effectiveness in improving reliability and reducing the cost of supervised mitigation.The paper "BEARS Make Neuro-Symbolic Models Aware of Their Reasoning Shortcuts" by Emanuele Marconato et al. addresses the issue of Reasoning Shortcuts (RSs) in Neuro-Symbolic (NeSy) models, which can lead to overconfidence and unreliable predictions. RSs occur when models learn concepts with unintended semantics, compromising reliability and generalization. The authors propose a method called BEARS (BE Aware of Reasoning Shortcuts) to make NeSy models aware of RSs by calibrating their concept-level confidence without sacrificing prediction accuracy. BEARS achieves this by ensembleizing concept extractors that are diverse in their predictions of concepts affected by RSs. Empirical results show that BEARS improves RS-awareness in state-of-the-art NeSy models and facilitates the acquisition of informative dense annotations for mitigation purposes. The paper also discusses the benefits and limitations of BEARS, highlighting its effectiveness in improving reliability and reducing the cost of supervised mitigation.