Diffusion Language Model Decoding

Papers on accelerating and improving inference in diffusion-based language models through techniques like speculative decoding, rejection sampling, and length-aware token generation to mitigate repetition and degeneration.