Non-Markovian Discrete Diffusion with Causal Language Models

TL;DR

This paper introduces CaDDi, a non-Markovian discrete diffusion model that conditions on entire sequences, enhancing expressive power and performance in language generation tasks compared to traditional Markovian diffusion models.

Contribution

CaDDi unifies causal and diffusion reasoning in a non-Markovian transformer, allowing for better sequence modeling and reuse of pretrained language model weights.

Findings

CaDDi outperforms existing discrete diffusion models on language benchmarks.

It narrows the performance gap between diffusion models and autoregressive transformers.

CaDDi can incorporate pretrained language models without architectural modifications.

Abstract

Discrete diffusion models offer a flexible, controllable approach to structured sequence generation, yet they still lag behind causal language models in expressive power. A key limitation lies in their reliance on the Markovian assumption, which restricts each step to condition only on the current state, leading to potential uncorrectable error accumulation. In this paper, we introduce CaDDi (Causal Discrete Diffusion Model), a discrete diffusion model that conditions on the entire generative trajectory, thereby lifting the Markov constraint and allowing the model to revisit and improve past states. By unifying sequential (causal) and temporal (diffusion) reasoning in a single non-Markovian transformer, CaDDi also treats standard causal language models as a special case and permits the direct reuse of pretrained LLM weights with no architectural changes. Empirically, CaDDi outperforms state-of-the-art discrete diffusion baselines on natural-language benchmarks, substantially narrowing the remaining gap to large autoregressive transformers.