Hi, I have a question regarding the use of publicly released pretrained models. The rules state that "publicly available data, including open pretrained networks" are allowed, but that "models pretrained on private or non-public data are not permitted." Could you clarify how this applies to publicly released foundation models where the checkpoint is openly available, but the original pretraining dataset is not fully released or downloadable? For example, would models such as DINOv2/DINOv3 (pretrained on large-scale datasets that are not publicly downloadable) or domain-specific models such as SurgeNetXL be allowed, provided that the pretrained checkpoint is publicly available and the model is properly declared? Thanks for clarifying!