open-sourcing the full Goedel-Prover-V2 training datasets for the community: SFT (1.74M samples) https://huggingface.co/datasets/Goedel-LM/SFT_dataset_v2 RL (whole proof generation + self-revision, 98k samples) https://huggingface.co/datasets/Goedel-LM/RL_dataset_V2