At the 13th International Conference on Computational Data and Social Networks (CSoNet 2024), Assoc. Prof. Quan Thanh Tho presented a groundbreaking paper titled “Energy-Based Learning for Robust Fake News Detection: A Graph Neural Network Approach with Trainable Cost Function”. The study introduces a novel approach to fake news detection by combining Graph Neural Networks (GNNs) with energy-based learning. Leveraging graph structures, the method aggregates data among interconnected nodes to capture meaningful relationships and dependencies – critical for identifying misinformation in complex networks. What sets this method apart is its innovative framing of the fake news detection task as an Out-of-Distribution (OOD) detection problem. By employing a trainable energy loss function, the model maximizes the energy gap between real and fake news, effectively enhancing robustness and accuracy. The proposed approach demonstrates performance comparable to state-of-the-art methods while showcasing the immense potential of energy-based models in combating fake news, a challenge of growing importance in today’s digital age. This study not only pushes the boundaries of fake news detection but also highlights the power of combining graph-based learning and energy models to tackle real-world problems.