UA-VLFM: An Uncertainty-aware Vision-Language Foundation Model for Auxiliary Diagnosis of Vitreoretinal Iymphoma | Synapse