Update README.md

Convert to SafeTensors format for security compliance

Upload fDPO trained Qwen3-0.6B model on MNLP M3 dataset (69k samples)

initial commit