LEAP: Unlocking dLLM Parallelism via Lookahead Early-Convergence Token Detection

TL;DR

LEAP is a novel method that detects early-converging tokens in diffusion language models, enabling faster parallel decoding without sacrificing accuracy by leveraging future context and multi-sequence superposition.

Contribution

LEAP introduces a training-free, plug-and-play technique for early token convergence detection, significantly reducing inference steps in dLLMs.

Findings

Reduces average denoising steps by about 30%.

Accelerates decoding to 7.2 tokens per step on GSM8K.

Enables reliable early decoding of tokens without high-confidence thresholds.

Abstract

Diffusion Language Models (dLLMs) have garnered significant attention for their potential in highly parallel processing. The parallel capabilities of existing dLLMs stem from the assumption of conditional independence at high confidence levels, which ensures negligible discrepancy between the marginal and joint distributions. However, the stringent confidence thresholds required to preserve accuracy severely constrain the scalability of parallelism. Through systematic token-level statistical analysis, we reveal that a substantial proportion of tokens converge to their correct predictions early in the denoising process yet fail to reach standard confidence thresholds, confirming that current confidence-based criteria are overly conservative. In response, we introduce LEAP (Lookahead Early-Convergence Token Detection for Accelerated Parallel Decoding). LEAP is a training-free, plug-and-play method that leverages future context filtering and multi-sequence superposition to detect early-converging tokens. By validating the alignment between early convergence and correctness, we enable reliable early decoding of these tokens. Benchmarking across diverse domains demonstrates that LEAP significantly lowers inference latency and decoding steps. Compared to confidence-based decoding, the average number of denoising steps is reduced by about 30%. On the GSM8K dataset, combining LEAP with dParallel accelerates decoding to 7.2 tokens per step while preserving model precision. LEAP effectively breaks the reliance on high-confidence priors, offering a novel paradigm for parallel decoding.