analiza aiqbench
This research introduces PRESTO, a framework that optimizes speculative decoding for diffusion large language models by addressing the inherent mismatch between parallel token generation and sequential verification processes. By implementing prefix-aligned scoring and priority-based tree search, the method significantly improves acceptance rates and throughput without compromising model accuracy. The work represents a critical advancement in making non-autoregressive architectures viable for high-speed inference pipelines.
PRESTO: Prefix-Aligned Tree Drafting for Diffusion Speculative Decoding
arXiv:2607.22634v1 Announce Type: new Abstract: Diffusion Large Language Models (dLLMs) have emerged as a promising alternative to autoregressive (AR) LLMs, generating tokens in parallel. This makes them effective draft models for speculative decoding (SD), producing an entire block of draft tokens in a single forward pass. Yet existing diffusion-based drafting methods rely on linear drafting, even though dLLMs emit multiple candidate tokens across positions, inducing a large combinatorial space of decoding paths. Consequently, they limit acceptance length and decoding efficiency. To exploit this multi-candidate structure, we apply tree-based drafting to diffusion drafters, enabling exploration of diverse candidate paths. However, we find that naive tree drafting is suboptimal: diffusion marginals are prefix-blind, mismatching the prefix-based AR verification and yielding unreliable path ranking. We propose PRESTO, a principled framework that extends tree-based drafting to diffusion drafters while resolving the fundamental mismatch between diffusion draft confidence and prefix-based AR verification through PREfix-aligned Scoring and priority-based Tree search for diffusion speculative decOding. The key principles behind PRESTO are that (1) candidate ranking should align with the prefix-based nature of AR verification, and (2) tree construction should prioritize candidate paths with high verification potential to maximize acceptance length. Extensive experiments show that PRESTO achieves up to an average of $1.5\times$ end-to-end throughput speedup on the state-of-the-art dedicated diffusion drafter SD and an average of $1.12\times$ on self-speculative diffusion LLMs across diverse benchmarks.
Abstract:Diffusion Large Language Models (dLLMs) have emerged as a promising alternative to autoregressive (AR) LLMs, generating tokens in parallel. This makes them effective draft models for speculative decoding (SD), producing an entire block of draft tokens in a single forward pass. Yet existing diffusion-based drafting methods rely on linear drafting, even though dLLMs emit multiple candidate tokens across positions, inducing a large combinatorial space of decoding paths. Consequently, they limit acceptance length and decoding efficiency. To exploit this multi-candidate structure, we apply tree-based drafting to diffusion drafters, enabling exploration of diverse candidate paths. However, we find that naive tree drafting is suboptimal: diffusion marginals are prefix-blind, mismatching the prefix-based AR verification and yielding unreliable path ranking. We propose PRESTO, a principled framework that extends tree-based drafting to diffusion drafters while resolving the fundamental mismatch between diffusion draft confidence and prefix-based AR verification through PREfix-aligned Scoring and priority-based Tree search for diffusion speculative decOding. The key principles behind PRESTO are that (1) candidate ranking should align with the prefix-based nature of AR verification, and (2) tree construction should prioritize candidate paths with high verification potential to maximize acceptance length. Extensive experiments show that PRESTO achieves up to an average of $1.5\times$ end-to-end throughput speedup on the state-of-the-art dedicated diffusion drafter SD and an average of $1.12\times$ on self-speculative diffusion LLMs across diverse benchmarks.
Artificial Intelligence (cs.AI)
arXiv:2607.22634 [cs.AI]
(or arXiv:2607.22634v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2607.22634 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 314 cuvinte
Subiecte conexe
JetSpec: Breaking the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting
arXiv · 30 sept. 2026
D^2SD: Accelerating Speculative Decoding with Dual Diffusion Draft Models
arXiv · 5 iun. 2026
[AINews] Agents disprove Polysemanticity; greatly improve Interpretability & Alignment
Latent Space · 16 oct. 2026
Is Multimodal Speculative Decoding Ready for Diffusion-Based Parallel Drafting? A Survey and Empirical Diagnosis
arXiv · 24 aug. 2026