Factorization-Error-Free Discrete Diffusion Language Model via Speculative Decoding

TL;DR

This paper introduces FeF-DLLM, a novel discrete diffusion language model that eliminates factorization errors and accelerates inference using speculative decoding, improving accuracy and speed.

Contribution

FeF-DLLM replaces independent token prediction with exact prefix-conditioned posterior factorization and incorporates speculative decoding for faster inference.

Findings

Improves accuracy by an average of 5.04 percentage points.

Achieves an average inference speedup of 3.86 times.

Generates from the true joint distribution, ensuring correctness.

Abstract

Discrete diffusion language models improve generation efficiency through parallel token prediction, but standard $X_0$ prediction methods introduce factorization errors by approximating the clean token posterior with independent token-wise distributions. This paper proposes Factorization-Error-Free Discrete Diffusion Language Modeling (FeF-DLLM), which replaces independent clean-token prediction with an exact prefix-conditioned factorization of the clean posterior to better preserve token dependencies. To reduce the sequential cost introduced by prefix conditioning, FeF-DLLM further incorporates speculative decoding within diffusion denoising, accelerating inference while maintaining the parallel prediction and re-masking properties of DLLMs. Theoretically, we prove that FeF-DLLM generates from the true joint distribution and derive its expected acceleration ratio. Experiments on GSM8K, MATH, HumanEval, and MBPP demonstrate that our method improves accuracy by an average of 5.04 percentage points while achieving an average inference speedup of $3.86\times$.