The field of neurosymbolic AI is witnessing significant advancements in logical reasoning and generation. Researchers are exploring innovative approaches to integrate symbolic learning with neural reasoning, enabling the development of more interpretable and reliable models. A key direction is the use of logical constraints to guide the generation of data, ensuring that the produced samples meet formal correctness criteria. This is achieved through the use of techniques such as Ehrenfeucht-Fraisse games and prototypical learning, which allow for more effective and efficient training of neurosymbolic models. Noteworthy papers in this area include: Logical GANs: Adversarial Learning through Ehrenfeucht Fraisse Games, which introduces a framework for logic-bounded generation with interpretable failures. Symbolic Neural Generation with Applications to Lead Discovery in Drug Design, which presents a hybrid neurosymbolic model for generating molecules that satisfy logical specifications. Right for the Right Reasons: Avoiding Reasoning Shortcuts via Prototypical Neurosymbolic AI, which proposes a prototypical neurosymbolic architecture to avoid reasoning shortcuts and learn correct basic concepts.