The paper introduces an energy-based framework for out-of-distribution (OOD) detection, addressing the issue of overconfident posterior distributions for OOD data using the softmax confidence score. The authors propose using an energy score, which is theoretically aligned with the probability density of inputs and is less prone to overconfidence. This framework can be used as a scoring function for pre-trained neural classifiers and as a trainable cost function to shape the energy surface explicitly for OOD detection. The energy score outperforms the softmax confidence score in distinguishing in- and out-of-distribution samples, reducing the average FPR by 18.03% on CIFAR-10. The method also outperforms state-of-the-art approaches on common benchmarks, demonstrating its effectiveness in both inference and training phases. The paper includes a comprehensive literature review and experimental results to support the proposed method's advantages over existing techniques.The paper introduces an energy-based framework for out-of-distribution (OOD) detection, addressing the issue of overconfident posterior distributions for OOD data using the softmax confidence score. The authors propose using an energy score, which is theoretically aligned with the probability density of inputs and is less prone to overconfidence. This framework can be used as a scoring function for pre-trained neural classifiers and as a trainable cost function to shape the energy surface explicitly for OOD detection. The energy score outperforms the softmax confidence score in distinguishing in- and out-of-distribution samples, reducing the average FPR by 18.03% on CIFAR-10. The method also outperforms state-of-the-art approaches on common benchmarks, demonstrating its effectiveness in both inference and training phases. The paper includes a comprehensive literature review and experimental results to support the proposed method's advantages over existing techniques.