Steering Without Breaking: Mechanistically Informed Interventions for Discrete Diffusion Language Models

ArXi:2605.10971v1 Announce Type: cross Discrete diffusion language models (DLMs) generate text by iteratively denoising all positions in parallel, offering an alternative to autoregressive models. Controlled generation methods for DLMs, imported from autoregressive models, apply uniform intervention at every denoising steps. We show this uniform schedule degrades quality, and the damage compounds when multiple attributes are steered jointly.