Machine Unlearning for Masked Diffusion Language Models

In three linesFirst machine unlearning framework for masked diffusion language models (LLaDA, Dream). MDU minimizes KL divergence from prompt-conditional to prompt-masked unconditional distribution at each masked position, with temperature scaling for privacy-utility trade-off. Code released.

Summary generated by Claude — human-verified