This study explores the role of coherence-based reasoning in belief updating within uncertain environments. We develop a novel computational model where agents update their beliefs based on observed evidence, with some evaluating the coherence of their belief set before accepting new evidence. Our results show that coherence-based evidence filtering improves belief accuracy in noisy environments and when agents' prior beliefs are accurate. However, when agents encounter systematically misleading evidence, coherence considerations lead to less accurate beliefs. These findings shed light on how coherence interacts with evidence quality and belief accuracy.