Hi, Following the thread "Training Data Policy for BraTS 2026 Challenge 1 — External Datasets and Pre-trained Weights", where the stated policy (no external datasets, no pre-trained weights) conflicted with the Challenge Rules T&C, and the discrepancy was acknowledged on 6/19 — the T&C page still reads permissively, so I'd like to confirm the position for Task 4. Specifically: for Task 4 (Inpainting), are participants permitted to use publicly released algorithms from previous BraTS inpainting challenges as components of a submission, without fine-tuning? The ambiguity is that the T&C prohibit pre-trained models "generated using previous BraTS challenge datasets." As I understand it — and the Challenge Rules note that "the inpainting challenge operates on the BraTS-GLI 2023 dataset" — algorithms from previous inpainting editions would have been trained on essentially the same publicly provided training data that 2026 participants receive, rather than on a different sub-challenge's dataset. Please correct me if that understanding is mistaken. A clear yes/no for Task 4 would be very helpful before finalizing containerization. Thank you.