Qing-Lin Zhu

Follow the Latent Roadmap: Navigating Revocable Decoding for Diffusion LLMs with Anchor Tokens

This work proposes ASRD (Anchor Supervised Revocable Decoding), a training-free framework that operates within the embedding space and outperforms recent remasking baselines, achieving accuracy improvements of up to 6.4\% while accelerating inference throughput by up to 7.2$\times$.