S2D2: Fast Decoding for Diffusion LLMs via Training-Free Self-Speculation

TL;DR

S2D2 introduces a training-free self-speculative decoding method for block-diffusion language models, enhancing speed and accuracy without extra training or test-time compute by combining diffusion with autoregressive verification.

Contribution

It proposes a novel hybrid decoding framework that uses the pretrained model as both drafter and verifier, improving speed and accuracy in block-diffusion models without additional training.

Findings

Up to 4.7× speedup over autoregressive decoding on SDAR

Up to 1.57× speedup over tuned dynamic decoding baseline

Accuracy improved by up to 4.5 points

Abstract

Block-diffusion language models offer a promising path toward faster-than-autoregressive generation by combining block-wise autoregressive decoding with within-block parallel denoising. However, in the few-step regime needed for practical acceleration, standard confidence-thresholded decoding is often brittle: aggressive thresholds hurt quality, while conservative thresholds require unnecessary denoising steps. Existing approaches that address this issue either require additional training or incur extra test-time compute. We present S2D2, a training-free self-speculative decoding framework for block-diffusion language models. Our key observation is that a block-diffusion model becomes autoregressive when the block size is reduced to one, allowing the same pretrained model to act as both drafter and verifier. S2D2 inserts a speculative verification step into standard block-diffusion decoding and uses lightweight routing policies to decide when verification is worth its cost. This yields a hybrid decoding trajectory in which diffusion proposes tokens in parallel, while the autoregressive mode acts as a local sequence-level critic. Across three mainstream block-diffusion families, S2D2 consistently improves the accuracy-speed tradeoff over strong confidence-thresholding baselines. On SDAR, we observe up to $4.7\times$ speedup over autoregressive decoding, and up to $1.57\times$ over a tuned dynamic decoding baseline while improving accuracy by up to $4.5$ points. On LLaDA2.1-Mini, S2D2 remains complementary to built-in self-correction, including a conservative setting where it is $4.4\times$ faster than the static baseline with slightly higher accuracy.