About Official repo of paper "SRUM: Fine-Grained Self-Rewarding for Unified Multimodal Models". A post-training framework that creates a cost-effective, self-iterative optimization loop. - View it on GitHub