NTH

Learning When to Commit: Schedule Distillation for Fast Diffusion Language Models

AuthorsAidar Myrzakhan, Zhiqiang Shen

AffiliationsVILA Lab, MBZUAI

November 10, 2026 2 min read
Watch on YouTube
The one-line take

dSpark teaches diffusion language models which tokens to reveal first, aiming to generate answers much faster without sacrificing quality.

Key results

5.1
Uncertain-token readiness gain

An uncertain reveal made 5.1× more other positions correct than a confident reveal.

4.7
Uncertain-token surprisal reduction

An uncertain reveal reduced other positions’ surprisal 4.7× more than a confident reveal.

10.12
LLaDA GSM8K decoding parallelism

dSpark achieved 10.12× the base model’s decoding parallelism.

5.12
Dream GSM8K decoding parallelism

dSpark achieved 5.12× the base model’s decoding parallelism.

249
LLaDA GSM8K throughput

Tokens per second on an NVIDIA A100; reported as 16.6× the base model’s throughput.

What the paper found

dSpark accelerates diffusion language models by learning when to reveal tokens, rather than simply committing whichever predictions are most confident. Its offline schedule construction promotes plausible tokens when revealing them reduces uncertainty across other masked positions; on GSM8K, revealing an uncertain token made 5.1× more other positions correct and reduced their surprisal 4.7× more than revealing a confident token. The team distills these schedules into LLaDA-8B-Instruct and Dream-7B-Instruct, using a KL-divergence penalty on failed trajectories to limit accuracy loss. Across GSM8K, MATH, HumanEval, and MBPP, dSpark reaches 10.12× the base model’s decoding parallelism on LLaDA and 5.12× on Dream for GSM8K, while LLaDA accuracy is 78.9%. On an NVIDIA A100, LLaDA generates 249 tokens per second on GSM8K, a reported 16.6× throughput gain over the base model. The central result is that useful context—not confidence alone—can guide faster parallel decoding while preserving task performance.

Original abstract

Diffusion large language models (dLLMs) can generate multiple tokens in parallel, but often require many decoding rounds to complete a response. Efficient generation depends not only on which tokens are easy to predict, but also on how committing to them helps predict the remaining positions. A highly predictable token may provide little useful context, while revealing a less predictable token can help several other positions become predictable sooner. We propose dSpark, a new schedule distillation-based framework for learning when to commit tokens for faster diffusion language generation. To achieve this, we construct unmasking schedules that prioritize tokens whose reveal makes other positions easier to predict, reducing uncertainty earlier in decoding. We then distill these schedules into the model to enable faster parallel sampling, with a KL-divergence regularizer to help preserve generation quality. We extensively evaluate our approach across mathematical reasoning and code generation tasks. On GSM8K, our method achieves remarkable 10.12$\times$ and 5.12$\times$ speedups over the various baselines on LLaDA-8B-Instruct and Dream-7B-Instruct, respectively, while maintaining superior task performance. Our code is available at https://github.com/VILA-Lab/dSpark.

Read the original paper

More in Diffusion Models

Browse all 82 papers →