The field of AI-driven scientific discovery is rapidly evolving, with a growing focus on developing formal epistemic systems and architectures that can promote truth utility, rational belief convergence, and audit-resilient integrity. Recent work has highlighted the importance of integrating cognitive and embodied intelligence to support iterative, autonomous experimentation and serendipitous discovery. There is also a increasing recognition of the need for more robust processes to support the systematic correction of errors and the refutation of flawed research. Furthermore, advances in natural language processing and information retrieval are enabling the development of more sophisticated scientific fact-checking systems and AI-driven review processes. Noteworthy papers in this area include: Bayesian Epistemology with Weighted Authority, which introduces a formally structured architecture for autonomous scientific reasoning; Active Inference AI Systems for Scientific Discovery, which outlines an architecture for AI-driven science that combines internal models with external validation; and The Next Phase of Scientific Fact-Checking, which examines the limitations of current scientific fact-checking systems and identifies key research challenges.