Cover Image

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

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Han, Ligong, Wang, Hao, Gao, Han, Xu, Kai, Srivastava, Akash
Format:	Preprint
Published:	2026
Subjects:	Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

SQuat: Subspace-orthogonal KV Cache Quantization
by: Wang, Hao, et al.
Published: (2025)

Speculate Deep and Accurate: Lossless and Training-Free Acceleration for Offloaded LLMs via Substitute Speculative Decoding
by: Wang, Pei-Shuo, et al.
Published: (2025)

Few-Step Diffusion Language Models via Trajectory Self-Distillation
by: Zhang, Tunyu, et al.
Published: (2026)

Self Speculative Decoding for Diffusion Large Language Models
by: Gao, Yifeng, et al.
Published: (2025)

Kangaroo: Lossless Self-Speculative Decoding via Double Early Exiting
by: Liu, Fangcheng, et al.
Published: (2024)

DART: Diffusion-Inspired Speculative Decoding for Fast LLM Inference
by: Liu, Fuliang, et al.
Published: (2026)

Goose: Anisotropic Speculation Trees for Training-Free Speculative Decoding
by: Jin, Tao, et al.
Published: (2026)

Training-Free Bayesianization for Low-Rank Adapters of Large Language Models
by: Shi, Haizhou, et al.
Published: (2024)

Fast Best-of-N Decoding via Speculative Rejection
by: Sun, Hanshi, et al.
Published: (2024)

PSD: Pushing the Pareto Frontier of Diffusion LLMs via Parallel Speculative Decoding
by: Sun, Shengyin, et al.
Published: (2026)

AdaSpec: Adaptive Speculative Decoding for Fast, SLO-Aware Large Language Model Serving
by: Huang, Kaiyu, et al.
Published: (2025)

Factorization-Error-Free Discrete Diffusion Language Model via Speculative Decoding
by: Fang, Xun, et al.
Published: (2026)

SNLP: Layer-Parallel Inference via Structured Newton Corrections
by: Han, Ligong, et al.
Published: (2026)

Speculating LLMs' Chinese Training Data Pollution from Their Tokens
by: Zhang, Qingjie, et al.
Published: (2025)

Dr. SoW: Density Ratio of Strong-over-weak LLMs for Reducing the Cost of Human Annotation in Preference Tuning
by: Xu, Guangxuan, et al.
Published: (2024)

SpeCache: Speculative Key-Value Caching for Efficient Generation of LLMs
by: Jie, Shibo, et al.
Published: (2025)

LogitSpec: Accelerating Retrieval-based Speculative Decoding via Next Next Token Speculation
by: Liu, Tianyu, et al.
Published: (2025)

EMS-SD: Efficient Multi-sample Speculative Decoding for Accelerating Large Language Models
by: Ni, Yunsheng, et al.
Published: (2024)

Training-Free Loosely Speculative Decoding: Accepting Semantically Correct Drafts Beyond Exact Match
by: Li, Jinze, et al.
Published: (2025)

Dynamic Depth Decoding: Faster Speculative Decoding for LLMs
by: Brown, Oscar, et al.
Published: (2024)

Fast Large Language Model Collaborative Decoding via Speculation
by: Fu, Jiale, et al.
Published: (2025)

Hopscotch: Discovering and Skipping Redundancies in Language Models
by: Eyceoz, Mustafa, et al.
Published: (2025)

Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding
by: Wu, Chengyue, et al.
Published: (2025)

DFlash: Block Diffusion for Flash Speculative Decoding
by: Chen, Jian, et al.
Published: (2026)

Temperature-Centric Investigation of Speculative Decoding with Knowledge Distillation
by: Ouyang, Siru, et al.
Published: (2024)

SpecDiff-2: Scaling Diffusion Drafter Alignment For Faster Speculative Decoding
by: Sandler, Jameson, et al.
Published: (2025)

Recurrent Drafter for Fast Speculative Decoding in Large Language Models
by: Cheng, Yunfei, et al.
Published: (2024)

Ouroboros: Generating Longer Drafts Phrase by Phrase for Faster Speculative Decoding
by: Zhao, Weilin, et al.
Published: (2024)

CLaSp: In-Context Layer Skip for Self-Speculative Decoding
by: Chen, Longze, et al.
Published: (2025)

RASD: Retrieval-Augmented Speculative Decoding
by: Quan, Guofeng, et al.
Published: (2025)

Streaming-dLLM: Accelerating Diffusion LLMs via Suffix Pruning and Dynamic Decoding
by: Xiao, Zhongyu, et al.
Published: (2026)

Draft & Verify: Lossless Large Language Model Acceleration via Self-Speculative Decoding
by: Zhang, Jun, et al.
Published: (2023)

Sculpting Subspaces: Constrained Full Fine-Tuning in LLMs for Continual Learning
by: Nayak, Nikhil Shivakumar, et al.
Published: (2025)

Cost-Aware Diffusion Draft Trees for Speculative Decoding
by: Zhang, Shuai, et al.
Published: (2026)

Accelerating Speculative Decoding with Block Diffusion Draft Trees
by: Ringel, Liran, et al.
Published: (2026)

Speculative Decoding with a Speculative Vocabulary
by: Williams, Miles, et al.
Published: (2026)

When Speculation Spills Secrets: Side Channels via Speculative Decoding In LLMs
by: Wei, Jiankun, et al.
Published: (2024)

Speculative Pipeline Decoding: Higher-Accruacy and Zero-Bubble Speculation via Pipeline Parallelism
by: Yu, Yijiong, et al.
Published: (2026)

SelfJudge: Faster Speculative Decoding via Self-Supervised Judge Verification
by: Yoon, Kanghoon, et al.
Published: (2025)

Mamba Drafters for Speculative Decoding
by: Choi, Daewon, et al.
Published: (2025)