Machine Unlearning: Qwen3-8B

what this is

Model on the left is an untouched Qwen3-8B; model on the right is the trained model with RMU + SFT + GRPO.

Source Writeup LinkedIn Twitter