CtrlDiff: Boosting Large Diffusion Language Models with Dynamic Block Prediction and Controllable Generation

TL;DR

CtrlDiff is a novel semi-autoregressive diffusion model that adaptively segments text for generation, improving controllability and efficiency while narrowing the performance gap with autoregressive models.

Contribution

It introduces a dynamic block prediction mechanism and a classifier-guided control method for diffusion-based language models, addressing fixed-length and controllability limitations.

Findings

Outperforms previous diffusion models in quality and controllability

Enables efficient post-hoc conditional generation without retraining

Reduces computational overhead compared to existing methods

Abstract

Although autoregressive models have dominated language modeling in recent years, there has been a growing interest in exploring alternative paradigms to the conventional next-token prediction framework. Diffusion-based language models have emerged as a compelling alternative due to their powerful parallel generation capabilities and inherent editability. However, these models are often constrained by fixed-length generation. A promising direction is to combine the strengths of both paradigms, segmenting sequences into blocks, modeling autoregressive dependencies across blocks while leveraging discrete diffusion to estimate the conditional distribution within each block given the preceding context. Nevertheless, their practical application is often hindered by two key limitations: rigid fixed-length outputs and a lack of flexible control mechanisms. In this work, we address the critical limitations of fixed granularity and weak controllability in current large diffusion language models. We propose CtrlDiff, a dynamic and controllable semi-autoregressive framework that adaptively determines the size of each generation block based on local semantics using reinforcement learning. Furthermore, we introduce a classifier-guided control mechanism tailored to discrete diffusion, which significantly reduces computational overhead while facilitating efficient post-hoc conditioning without retraining. Extensive experiments demonstrate that CtrlDiff sets a new standard among hybrid diffusion models, narrows the performance gap to state-of-the-art autoregressive approaches, and enables effective conditional text generation across diverse tasks.