Wide-In, Narrow-Out: Revokable Decoding for Efficient and Effective DLLMs

TL;DR

The paper introduces WINO, a decoding algorithm for DLLMs that enhances speed and quality by enabling revokable, parallel token generation and verification, addressing the irreversibility issue in standard decoding.

Contribution

WINO is a training-free, parallel decoding method that improves DLLM performance by allowing token verification and refinement during inference.

Findings

Accelerates inference by 6× on GSM8K with 2.58% accuracy gain

Achieves 10× speedup on Flickr30K with better performance

Demonstrates superiority over existing decoding methods

Abstract

Diffusion Large Language Models (DLLMs) have emerged as a compelling alternative to Autoregressive models, designed for fast parallel generation. However, existing DLLMs are plagued by a severe quality-speed trade-off, where faster parallel decoding leads to significant performance degradation. We attribute this to the irreversibility of standard decoding in DLLMs, which is easily polarized into the wrong decoding direction along with early error context accumulation. To resolve this, we introduce Wide-In, Narrow-Out (WINO), a training-free decoding algorithm that enables revokable decoding in DLLMs. WINO employs a parallel draft-and-verify mechanism, aggressively drafting multiple tokens while simultaneously using the model's bidirectional context to verify and re-mask suspicious ones for refinement. Verified in open-source DLLMs like LLaDA and MMaDA, WINO is shown to decisively improve the quality-speed trade-off. For instance, on the GSM8K math benchmark, it accelerates inference by 6$\times$ while improving accuracy by 2.58%; on Flickr30K captioning, it achieves a 10$\times$ speedup with higher performance. More comprehensive experiments are conducted to demonstrate the superiority and provide an in-depth understanding of WINO.