Datasets:

fineset-io
/

speculative-decoding-papers

id string	sources list	title string	abstract string	authors list	categories list	fields_of_study list	published_date timestamp[s]	url string	pdf_url string	arxiv_id string	doi string	citation_count int64	has_code bool	code_url string	venue string	quality_score float64
989fef5fc253b0613a59b01cec35c0cc584fc413ce901e3b5192ea8bd456c69e	[ "arxiv" ]	SSD: Spatially Speculative Decoding Accelerates Autoregressive Image Generation	Autoregressive models excel in visual generation by treating images as 1D sequences of discrete tokens, mirroring language modeling. However, this flattening discards the intrinsic 2D spatial locality of visual signals, creating severe computational bottlenecks during inference. We introduce Spatially Speculative Decod...	[ "Shilong Xiang", "Zirui Zhang", "Lijun Yu", "Chengzhi Mao" ]	[ "cs.CV" ]	[]	2026-06-18T00:00:00	https://arxiv.org/abs/2606.20543	https://arxiv.org/pdf/2606.20543v1	2606.20543	null	0	false	null	null	0.35
d6953c2d8a4c8fbbfaf4feaf13b58c3bca43fa126b3c3ad1191d86913ae7b0e1	[ "arxiv" ]	EfficientRollout: System-Aware Self-Speculative Decoding for RL Rollouts	Reinforcement learning (RL) has become a representative post-training paradigm for LLMs, enabling strong reasoning and agentic capabilities. However, rollout generation remains a dominant latency bottleneck because autoregressive sampling decodes responses sequentially and a small number of long-tailed generations ofte...	[ "Minseo Kim", "Minjae Lee", "Seunghyuk Oh", "Kevin Galim", "Donghoon Kim", "Coleman Hooper", "Harman Singh", "Amir Gholami", "Hyung Il Koo", "Wonjun Kang" ]	[ "cs.LG" ]	[]	2026-06-17T00:00:00	https://arxiv.org/abs/2606.18967	https://arxiv.org/pdf/2606.18967v1	2606.18967	null	0	true	https://github.com/furiosa-ai/EfficientRollout	null	0.65
e6b1bda0dd2caa8573a59a60e8a011d3ac98fec3b34a3f58fa4f3c9d97fba360	[ "arxiv" ]	JetFlow: Breaking the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting	Speculative decoding (SD) accelerates autoregressive Large Language Models (LLMs) by drafting multiple tokens and verifying them in parallel, but it faces a scaling limitation: increasing the draft budget improves speed only when acceptance remains high and drafting overhead stays low. This ceiling has been difficult t...	[ "Lanxiang Hu", "Zhaoxiang Feng", "Yulun Wu", "Haoran Yuan", "Yujie Zhao", "Yu-Yang Qian", "Bojun Wang", "Daxin Jiang", "Yibo Zhu", "Tajana Rosing", "Hao Zhang" ]	[ "cs.CL" ]	[]	2026-06-16T00:00:00	https://arxiv.org/abs/2606.18394	https://arxiv.org/pdf/2606.18394v1	2606.18394	null	0	true	https://github.com/hao-ai-lab/JetFlow	null	0.65
e7d194915f2cea21b85eb67bb8325e9557465399c70a6708c0c19dfea7a2b985	[ "arxiv", "semantic_scholar" ]	Accelerating Speculative Diffusions via Block Verification	Speculative decoding speeds up LLM inference by using a draft model to generate tokens, with an acceptance-rejection scheme that ensures that the output matches the target distribution. Adapting this to continuous diffusions is difficult because speculative sampling requires drawing from a residual distribution. While ...	[ "Alexander Soen", "Hisham Husain", "Valentin De Bortoli", "Arnaud Doucet" ]	[ "cs.LG", "stat.ML" ]	[ "Computer Science", "Mathematics" ]	2026-06-11T00:00:00	https://arxiv.org/abs/2606.13426	https://arxiv.org/pdf/2606.13426v1	2606.13426	null	0	false	null	null	0.35
9835e0ad1804d5b57d689525445d89ace6b367cb70c1fcd0dc5e0ca3918f2874	[ "arxiv", "semantic_scholar" ]	VIA-SD: Verification via Intra-Model Routing for Speculative Decoding	Speculative decoding (SD) addresses the high inference costs of LLMs by having lightweight drafters generate candidates for large verifiers to validate in parallel. Existing draft-verify methods use binary decisions: accept or fully recompute. Yet we find that many rejected tokens can be verified correctly by a slim su...	[ "Yuchen Xian", "Yang He", "Yunqiu Xu", "Yi Yang" ]	[ "cs.CL", "cs.AI" ]	[ "Computer Science" ]	2026-06-10T00:00:00	https://arxiv.org/abs/2606.12243	https://arxiv.org/pdf/2606.12243v1	2606.12243	null	0	false	null	null	0.35
0575977791d2291d5bbfd253a988f4ac64d1c8e3d8206b49af1ca202bb46201b	[ "arxiv", "semantic_scholar" ]	PathRelax: Parallel-Path Relaxed Speculative Jacobi Decoding for Accelerating Auto-Regressive Text-to-Image Generation	The growing need for high-resolution image generation in autoregressive text-to-image models has resulted in extended token sequences, significantly increasing computational costs and inference times. However, existing state-of-the-art methods for accelerating autoregressive text-to-image models rely on chain-structure...	[ "Haodong Lei", "Hongsong Wang", "Bingxuan Dai", "Pan Zhou" ]	[ "cs.CV" ]	[ "Computer Science" ]	2026-06-09T00:00:00	https://arxiv.org/abs/2606.10492	https://arxiv.org/pdf/2606.10492v1	2606.10492	null	0	true	https://github.com/Haodong-Lei-Ray/PathSpec	null	0.65
6ced2fae8c1c2631d3c6e33145f50e7d59272f75d55da532305c73c60ee63c94	[ "arxiv", "semantic_scholar" ]	TRADE: Transducer-Augmented Decoder for Speech LLM	Speech Large Language Models (Speech LLMs) lack a principled mechanism for streaming inference: their label-synchronous generation has no acoustic-frame alignment, making real-time decoding and end-of-utterance detection difficult. We propose TRADE TRansducer-Augmented DEcoder, which augments a multimodal LLM with a tr...	[ "Yun Tang", "Shanil Puri", "Shinji Watanabe", "Subhabrata Mukherjee" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-06-07T00:00:00	https://arxiv.org/abs/2606.08486	https://arxiv.org/pdf/2606.08486v1	2606.08486	null	0	false	null	null	0.35
99301d0dea92075618a6928d093fa57ef75c64b427ac22f38fce006a762521c4	[ "arxiv", "semantic_scholar" ]	WhiFlash: Accelerating Speculative Decoding with Token-Level Cross-Paradigm Routing	The autoregressive nature of large language models (LLMs) remains a significant bottleneck for inference, particularly in complex agentic workloads. While speculative decoding (SD) accelerates inference, current approaches rely on static drafting paradigms, utilising either autoregressive drafting models for reasoning ...	[ "Young D. Kwon", "Miles Williams", "Rui Li", "Alexandros Kouris", "Stylianos I. Venieris" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-06-05T00:00:00	https://arxiv.org/abs/2606.07710	https://arxiv.org/pdf/2606.07710v1	2606.07710	null	0	false	null	null	0.35
5ea3182923a447700ff7de9a75dc9d912a92dbc16a5d4fdefccb458d42d9e167	[ "arxiv", "semantic_scholar" ]	YouZhi: Towards High-Concurrency Financial LLMs via Adaptive GQA-to-MLA Transition	Large language models (LLMs) drive significant financial innovations, yet their high-concurrency deployment is severely bottlenecked by KV cache memory overhead, which inflates infrastructure costs and throttles scalability. To address this, we propose YouZhi-LLM, a highly efficient financial LLM empowered by a compreh...	[ " PSBC LLM Team", " Huawei LLM Team", "Ruihan Long", "Junjie Wu", "Tianan Zhang", "Duo Zhang", "Yaozong Wu", "Jinbin Fu", "Chang Liu", "Zhentao Tang", "Wenshuang Yang", "Xin Wang", "Zhihao Song", "Ning Huang", "Wenjing Xu", "Shuai Zong", "Shupei Sun", "Sen Wang", "Jing Hu", "Bi...	[ "cs.CL" ]	[ "Computer Science" ]	2026-06-04T00:00:00	https://arxiv.org/abs/2606.05868	https://arxiv.org/pdf/2606.05868v1	2606.05868	null	0	false	null	null	0.35
0d4ec2a1e770ac73e7ea7b4d3e239d8bcf1fb1c16fc4ebfdbbffb26fc691afd3	[ "arxiv", "semantic_scholar" ]	AdaPLD: Adaptive Retrieval and Reuse for Efficient Model-Free Speculative Decoding	Speculative decoding accelerates generation by verifying multiple drafted tokens in a single target-model forward pass, reducing sequential decoding iterations. Model-free variants avoid auxiliary draft models by reusing text and model states already available during generation, but their speedup depends on the reliabi...	[ "Runheng Liu", "Jincheng Xie", "Wen Hu", "Xingchen Xiao", "Heyan Huang" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-06-04T00:00:00	https://arxiv.org/abs/2606.05742	https://arxiv.org/pdf/2606.05742v2	2606.05742	null	0	false	null	null	0.35
b59b2e4af5fd17912a3853612cec760e54c9bf310adf06402a54003a30186ded	[ "arxiv", "semantic_scholar" ]	D^2SD: Accelerating Speculative Decoding with Dual Diffusion Draft Models	Speculative decoding accelerates autoregressive large language model inference by drafting multiple tokens and verifying them in a single target-model forward pass. Recent diffusion-based drafters generate an entire block of tokens in parallel but usually commit to a single draft sequence per verification: once the fir...	[ "Liyuan Zhang", "Jiarui Zhang", "Jinwei Yao", "Ran Yan", "Yuchen Yang", "Jiahao Zhang", "Tongkai Yang", "Yi Wu", "Binhang Yuan" ]	[ "cs.DC", "cs.LG" ]	[ "Computer Science" ]	2026-06-03T00:00:00	https://arxiv.org/abs/2606.04446	https://arxiv.org/pdf/2606.04446v1	2606.04446	null	0	false	null	null	0.35
7bfd76541cf7094864c120ea7c6ab296bf68599cade499b5e32c163a38559bc0	[ "arxiv", "semantic_scholar" ]	TreeFlash: Parallel AR-Approximation for Faster Speculative Decoding	One-shot block drafters for speculative decoding generate the full draft in a single forward pass, achieving strong throughput by eliminating sequential token generation. However, they predict each draft token conditioned only on the prefix context, with no dependence on previously drafted tokens. This non-autoregressi...	[ "Peer Rheinboldt", "Frédéric Berdoz", "Roger Wattenhofer" ]	[ "cs.LG" ]	[ "Computer Science" ]	2026-06-02T00:00:00	https://arxiv.org/abs/2606.03819	https://arxiv.org/pdf/2606.03819v1	2606.03819	null	0	false	null	null	0.35
a4d7172a68c2513a2ea474b2fd98d98799697e94d04d32097615ad79a8bb5bff	[ "arxiv", "semantic_scholar" ]	SimSD: Simple Speculative Decoding in Diffusion Language Models	Diffusion large language models (dLLMs) have recently emerged as a promising alternative to autoregressive (AR) LLMs, offering faster inference through parallel or blockwise decoding. However, their masked language modeling formulation remains incompatible with standard token-level speculative decoding, one of the most...	[ "Junxia Cui", "Haotian Ye", "Runchu Tian", "Hongcan Guo", "Jinya Jiang", "Haoru Li", "Chaojie Ren", "Yiming Huang", "Kaijie Zhu", "Zhongkai Yu", "Kun Zhou", "Jingbo Shang" ]	[ "cs.CL", "cs.AI" ]	[ "Computer Science" ]	2026-06-01T00:00:00	https://arxiv.org/abs/2606.02544	https://arxiv.org/pdf/2606.02544v1	2606.02544	null	0	true	https://github.com/airevo2/SimSD-release	null	0.65
ce69ae40ce9fed8ab2bc6e71c105119c207ec66e9a30dcb3c5bf93b5a3d25ccb	[ "arxiv", "semantic_scholar" ]	Fast-dLLM++: Fréchet Profile Decoding for Faster Diffusion LLM Inference	Diffusion large language models promise parallel token generation, yet inference remains bottlenecked by deciding which masked tokens can be safely committed together. Fast-dLLM addressed this with KV caching and confidence-guided parallel decoding, but its decoding theory uses a homogeneous high-confidence assumption ...	[ "Siva Rajesh Kasa", "Yasong Dai", "Sumit Negi", "Hongdong Li" ]	[ "cs.CL", "cs.AI", "cs.LG" ]	[ "Computer Science" ]	2026-06-01T00:00:00	https://arxiv.org/abs/2606.02955	https://arxiv.org/pdf/2606.02955v2	2606.02955	null	0	true	https://github.com/Ringo-Star/FastdLLM_plusplus	null	0.65
ff9c8a7f73a37a9c55cb13592f6c9a8327401cb38f1c1d2844082cddc3ab203d	[ "arxiv", "semantic_scholar" ]	DFlare: Scaling Up Draft Capacity for Block Diffusion Speculative Decoding	Block diffusion speculative decoding accelerates LLM inference by predicting all tokens within a block simultaneously for the target model to verify in parallel. Predicting an entire block at once requires a sufficiently capable draft model and effective utilization of the target model's internal knowledge. However, th...	[ "Jiebin Zhang", "Zhenghan Yu", "Song Liu", "Eugene J. Yu", "Zheng Li", "Dawei Zhu", "Jiangshan Duo", "Weimin Xiong", "Yifan Song", "Guanghua Yu", "Jianchen Zhu", "Sujian Li" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-06-01T00:00:00	https://arxiv.org/abs/2606.02091	https://arxiv.org/pdf/2606.02091v2	2606.02091	null	0	true	https://github.com/Tencent/AngelSlim	null	0.65
67cb850d1ae2bbad9e335769689abaedd7c3f713ce53eb560d69f14e04724a4c	[ "arxiv", "semantic_scholar" ]	Cost-Aware Diffusion Draft Trees for Speculative Decoding	Speculative decoding accelerates inference by having a lightweight drafter propose tokens verified in parallel by the target language model. Block diffusion drafters such as DFlash generate an entire draft block in one pass, yielding per-position marginals; DDTree uses these to build a candidate tree that maximizes exp...	[ "Shuai Zhang", "Huachuan Qiu", "Hongliang He", "Yong Dai" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-06-01T00:00:00	https://arxiv.org/abs/2606.01813	https://arxiv.org/pdf/2606.01813v1	2606.01813	null	0	false	null	null	0.35
9aaa22aa6ff1b729f41b82c2327c674ae94b8d7e4fa885340d0b12ae2283d19d	[ "arxiv", "semantic_scholar" ]	Hybrid Verified Decoding: Learning to Allocate Verification in Speculative Decoding	Large Language Model (LLM) generation remains expensive because autoregressive decoding calls the model once for each new token. Speculative decoding reduces this cost by drafting multiple tokens and verifying them with the target model in one step, but its speedup depends on how many drafted tokens are accepted. Param...	[ "Xin Su", "Dawid Majchrowski", "Fangyuan Yu", "Vanshil Atul Shah", "Sebastian Rogawski", "Pawel Morkisz", "Anahita Bhiwandiwalla", "Phillip Howard" ]	[ "cs.CL", "cs.AI" ]	[ "Computer Science" ]	2026-05-31T00:00:00	https://arxiv.org/abs/2606.01019	https://arxiv.org/pdf/2606.01019v1	2606.01019	null	0	false	null	null	0.35
9dcd3f203797ef30a7139152faf9fdacb037877a4e76d5d63f75f843665a6110	[ "arxiv", "semantic_scholar" ]	DREAM-S: Speculative Decoding with Searchable Drafting and Target-Aware Refinement for Multimodal Generation	Speculative decoding (SD) has proven to be an effective technique for accelerating autoregressive generation in large language models (LLMs) however, its application to vision-language models (VLMs) remains relatively unexplored. We propose~\textit{DREAM-S}, a novel SD framework designed specifically for fast and effic...	[ "Zining Liu", "Yunhai Hu", "Tianhua Xia", "Bo Bao", "Eric Sather", "Vithursan Thangarasa", "Sai Qian Zhang" ]	[ "cs.LG" ]	[ "Computer Science" ]	2026-05-30T00:00:00	https://arxiv.org/abs/2606.00535	https://arxiv.org/pdf/2606.00535v1	2606.00535	null	0	true	https://github.com/SAI-Lab-NYU/DREAM-S	null	0.65
7b5bf44305ea75c39b6a396737e311e99fa30b92251fc32e114e68b2d9a10252	[ "arxiv", "semantic_scholar" ]	Doing What They Say, Not What They Reason: Locating the Faithfulness Gap in LLM Agents	Do LLM agents act on the reasoning they state? This question of process fidelity is central to using LLMs in social simulation, yet it is hard to measure where no reference for correct behavior exists. We study it in acontrolled setting, a Texas Poker simulator with a verifiable reference action for every decision by d...	[ "Yufeng Wang" ]	[ "cs.AI" ]	[ "Computer Science" ]	2026-05-30T00:00:00	https://arxiv.org/abs/2606.00476	https://arxiv.org/pdf/2606.00476v1	2606.00476	null	0	false	null	null	0.35
6cd545696b973a18c08ac72a12a584dc716d804210a521d7a889c32871f8e2a8	[ "arxiv", "semantic_scholar" ]	TAPS: Target-Aware Prefix Tree Selection for Diffusion-Drafted Speculative Decoding	Using a diffusion model for parallel drafting is a promising approach for speculative decoding. By predicting tokens at multiple future positions in a single forward pass, diffusion drafters substantially reduce drafting latency. However, this shifts the bottleneck to verification: verifying a single sequence limits ac...	[ "Zhuoyu Wang", "Junnan Huang", "Xinyu Chen" ]	[ "cs.AI" ]	[ "Computer Science" ]	2026-05-30T00:00:00	https://arxiv.org/abs/2606.00487	https://arxiv.org/pdf/2606.00487v1	2606.00487	null	0	false	null	null	0.35
9f4328520b446d19330d7c4dcbccaf6bd9dbbfc74e444eea7c61e2318d7fb401	[ "arxiv", "semantic_scholar" ]	Speculative Pipeline Decoding: Higher-Accruacy and Zero-Bubble Speculation via Pipeline Parallelism	Speculative Decoding (SD) accelerates low-concurrency LLM inference by employing a draft-then-verify paradigm. However, mainstream methods typically rely on multi-token prediction, which introduces escalating prediction difficulty and serial drafting latency. To address these, we propose Speculative Pipeline Decoding (...	[ "Yijiong Yu", "Huazheng Wang", "Shuai Yuan", "Ruilong Ren", "Ji Pei" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-05-29T00:00:00	https://arxiv.org/abs/2605.30852	https://arxiv.org/pdf/2605.30852v1	2605.30852	null	0	true	https://github.com/yuyijiong/speculative_pipeline_decoding	null	0.65
8f8355730b9437b1a8ad4f174b4109120e3f6c3a11321f06f206f5ee715c2ba2	[ "arxiv", "semantic_scholar" ]	BudgetDraft: Acceptance-Aware Multi-View Training for Sparse-KV Speculative Decoding	Speculative decoding speeds up autoregressive decoding by using a drafter to propose multiple tokens that a verifier validates in parallel. In resource-constrained deployments, the drafter uses a sparse KV cache to limit peak GPU memory and end-to-end latency under a fixed KV budget, while the verifier keeps a full KV ...	[ "Liang He", "Jingbo Wen", "Qishi Zhan", "Yixiong Chen", "Kangning Cui", "Qizhen Lan", "Xilu Wang" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-05-29T00:00:00	https://arxiv.org/abs/2606.00144	https://arxiv.org/pdf/2606.00144v1	2606.00144	null	0	false	null	null	0.35
6b7e8f9688dd54d8424545515380d722a0edb32b810ce0c2c9343e6299b7882f	[ "arxiv", "semantic_scholar" ]	Speculative Decoding Across Languages	Speculative decoding has become a crucial component of large language model (LLM) inference, enabling faster generation by drafting multiple tokens and verifying them in parallel. However, small draft models tend to suffer from disproportionately poor multilingual capabilities. Thus, when generating text in a non-Engli...	[ "Nirajan Paudel", "Michael Ginn", "Luc De Nardi", "Alexis Palmer" ]	[ "cs.CL", "cs.LG" ]	[ "Computer Science" ]	2026-05-28T00:00:00	https://arxiv.org/abs/2605.30580	https://arxiv.org/pdf/2605.30580v1	2605.30580	null	0	false	null	null	0.35
5e8bc4dba4e0946c2cf35ce10d096d6e2fdb902234a5ca8311d8059988d8e206	[ "arxiv", "semantic_scholar" ]	Domino: Decoupling Causal Modeling from Autoregressive Drafting in Speculative Decoding	Speculative decoding accelerates LLM inference by drafting multiple tokens and verifying them in parallel with the target model. However, its practical speedup is constrained by the trade-off between draft quality and drafting cost: autoregressive drafters model causal dependencies among draft tokens but incur sequenti...	[ "Jianuo Huang", "Yaojie Zhang", "Qituan Zhang", "Hao Lin", "Hanlin Xu", "Linfeng Zhang" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-05-28T00:00:00	https://arxiv.org/abs/2605.29707	https://arxiv.org/pdf/2605.29707v1	2605.29707	null	0	false	null	null	0.35
147639c70b8115768cc73c5f632a6d7d2eb13266393202b58718d927c21f38c3	[ "arxiv", "semantic_scholar" ]	Bastion: Budget-Aware Speculative Decoding with Tree-structured Block Diffusion Drafting	Block-diffusion drafters have recently emerged as a powerful alternative for speculative decoding by predicting multiple future-token distributions in a single parallel step. However, since these parallel predictions are sampled from position-wise marginals rather than fully conditioned sequences, committing to a singl...	[ "Soowon Oh", "Nam Cao", "Yujin Kim", "Hojung Jung", "Huzama Ahmad", "Sangmin Bae", "Se-Young Yun" ]	[ "cs.LG" ]	[ "Computer Science" ]	2026-05-28T00:00:00	https://arxiv.org/abs/2605.29727	https://arxiv.org/pdf/2605.29727v1	2605.29727	null	0	false	null	null	0.35
be70dc39d496949d095aaa790f0119c32b195a8e0989576371f7628e77995504	[ "arxiv", "semantic_scholar" ]	Cassandra: Enabling Reasoning LLMs at Edge via Self-Speculative Decoding	Speculative decoding has emerged as a promising lossless approach for accelerating Large Language Models (LLMs). As reasoning LLMs increasingly suffer from decode-stage overhead and approximation-based methods degrade accuracy, lossless speculative decoding has become essential for efficient inference. However, existin...	[ "Soongyu Choi", "Yuntae Kim", "Muyoung Son", "Joo-Young Kim" ]	[ "cs.AR" ]	[ "Computer Science" ]	2026-05-26T00:00:00	https://arxiv.org/abs/2605.26558	https://arxiv.org/pdf/2605.26558v1	2605.26558	null	0	false	null	null	0.35
416cf4347e2e0bec6775e27965ba73091905b9d5ef5ebdad4c9f6707001619a9	[ "arxiv", "semantic_scholar" ]	Beyond the Target: From Imitation to Collaboration in Speculative Decoding	Speculative decoding (SPD) accelerates large language model (LLM) inference by letting a smaller draft model propose multiple future tokens that are verified in parallel by a larger target model. The dominant SPD paradigm treats the target model as the sole reliable teacher, accepting a draft token only when it exactly...	[ "Jinze Li", "Yixing Xu", "Guanchen Li", "Jinfeng Xu", "Shuo Yang", "Yang Zhang", "Xuanwu Yin", "Dong Li", "Edith C. H. Ngai", "Emad Barsoum" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-05-24T00:00:00	https://arxiv.org/abs/2605.24793	https://arxiv.org/pdf/2605.24793v1	2605.24793	null	0	false	null	null	0.35
0a2277aa69ce5c776cfaf8fbe3fc063c6e90bc08066b695bd32a8a5a3699b43e	[ "arxiv", "semantic_scholar" ]	Optimus: Elastic Decoding for Efficient Diffusion LLM Serving	Large language model (LLM) serving is fundamentally limited by inefficient hardware utilization. Autoregressive (AR) decoding underutilizes GPUs due to its strictly sequential execution, while diffusion LLMs (DLLMs) improve throughput by decoding multiple tokens per iteration. However, fixed block-size diffusion decodi...	[ "Chiyue Wei", "Cong Guo", "Bowen Duan", "Junyao Zhang", "Haoxuan Shan", "Yifei Wang", "Yangjie Zhou", "Hai \"Helen\" Li", "Danyang Zhuo", "Yiran Chen" ]	[ "cs.DC" ]	[ "Computer Science" ]	2026-05-24T00:00:00	https://arxiv.org/abs/2605.24832	https://arxiv.org/pdf/2605.24832v1	2605.24832	null	0	true	https://github.com/dubcyfor3/Optimus	null	0.65
098017fe4a68a6d123d033c48f94c0761c518f8591dda3385cd29a2948cd4038	[ "arxiv", "semantic_scholar" ]	FlexDraft: Flexible Speculative Decoding via Attention Tuning and Bonus-Guided Calibration	Speculative decoding accelerates memory-bound LLM inference without quality degradation by using a fast drafter to propose multiple candidate tokens and the target model to verify them in parallel. However, conventional sequential speculative decoding suffers from mutual waiting between drafting and verification, and r...	[ "Yaojie Zhang", "Jianuo Huang", "Junlong Ke", "Yuhang Han", "Yongji Long", "Tianchen Zhao", "Biqing Qi", "Linfeng Zhang" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-05-19T00:00:00	https://arxiv.org/abs/2605.20022	https://arxiv.org/pdf/2605.20022v1	2605.20022	null	0	false	null	null	0.35
ebd62303ca3680e0604a8750d46aa645cf905886d42d9538aefdf0ed61443223	[ "arxiv", "semantic_scholar" ]	SSV: Sparse Speculative Verification for Efficient LLM Inference	Speculative decoding and dynamic sparse attention are two complementary approaches for accelerating long-context LLM inference: the former amortizes target-model execution across multiple verifier queries, while the latter reduces each query's KV-cache working set. Directly combining them, however, exposes a structural...	[ "Zhibin Wang", "Ziyu Zhong", "Nuo Shen", "Yuhang Zhou", "Rong Gu", "Sheng Zhong" ]	[ "cs.OS" ]	[ "Computer Science" ]	2026-05-19T00:00:00	https://arxiv.org/abs/2605.19893	https://arxiv.org/pdf/2605.19893v2	2605.19893	null	0	false	null	null	0.35
c14a952a4a42efb330843fb7d5cdae09ff8aa24c1ce0c2ea16595a1712a5037c	[ "arxiv", "semantic_scholar" ]	Mix-Quant: Quantized Prefilling, Precise Decoding for Agentic LLMs	LLM agents have recently emerged as a powerful paradigm for solving complex tasks through planning, tool use, memory retrieval, and multi-step interaction. However, these agentic workflows often introduce substantial input-side overhead, making the compute-intensive prefilling stage a key bottleneck in long-context, mu...	[ "Haiquan Lu", "Zigeng Chen", "Gongfan Fang", "Xinyin Ma", "Xinchao Wang" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-05-19T00:00:00	https://arxiv.org/abs/2605.20315	https://arxiv.org/pdf/2605.20315v1	2605.20315	null	0	false	null	null	0.35
4be3d37b8d8128e8ae64c55f0f86437c437d7f2663f224e7048d640def9c6757	[ "arxiv", "semantic_scholar" ]	Lever: Speculative LLM Inference on Smartphones	Large language models (LLMs) are increasingly needed for interactive mobile applications, but high-quality models exceed the limited DRAM available on smartphones. Flash storage can hold larger models, yet flash-backed inference is slow because autoregressive decoding repeatedly invokes the target model and incurs cost...	[ "Tuowei Wang", "Fengzu Li", "Yanfan Sun", "Wei Gao", "Ju Ren" ]	[ "cs.LG" ]	[ "Computer Science" ]	2026-05-16T00:00:00	https://arxiv.org/abs/2605.16786	https://arxiv.org/pdf/2605.16786v1	2605.16786	null	0	false	null	null	0.35
e09688cdb4c0604a4d4ca4a4cdcdfb445f9347986546c5bd91ea9917c7b82b8d	[ "arxiv", "semantic_scholar" ]	PSD: Pushing the Pareto Frontier of Diffusion LLMs via Parallel Speculative Decoding	Diffusion large language models (dLLMs) generate text by iteratively denoising masked token sequences. Although dLLMs can predict all masked positions in parallel within each step, the large number of denoising iterations still makes inference expensive. This cost can be reduced spatially by unmasking multiple tokens p...	[ "Shengyin Sun", "Yiming Li", "Renxi Liu", "Xinqi Li", "Hui-Ling Zhen", "Weizhe Lin", "Chen Chen", "Xianzhi Yu", "Mingxuan Yuan", "Chen Ma" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-05-15T00:00:00	https://arxiv.org/abs/2605.15609	https://arxiv.org/pdf/2605.15609v1	2605.15609	null	0	false	null	null	0.35
6233bae9530757780e00012ac3b436f2dfb46e51647691339f185f6928c4d269	[ "arxiv", "semantic_scholar" ]	An Interpretable Latency Model for Speculative Decoding in LLM Serving	Speculative decoding (SD) accelerates large language model (LLM) inference by using a smaller draft model to propose multiple tokens that are verified by a larger target model in parallel. While prior work demonstrates substantial speedups in isolated or fixed-batch settings, the behavior of SD in production serving sy...	[ "Linghao Kong", "Megan Flynn", "Michael Peng", "Nir Shavit", "Mark Kurtz", "Alexandre Marques" ]	[ "cs.LG", "cs.PF" ]	[ "Computer Science" ]	2026-05-14T00:00:00	https://arxiv.org/abs/2605.15051	https://arxiv.org/pdf/2605.15051v1	2605.15051	null	0	false	null	null	0.35
e76ad8097ba449fe5930b44fe71d728fffdbd7c97ee1f9462fcc8e0dd934a78a	[ "arxiv", "semantic_scholar" ]	Performance-Driven Policy Optimization for Speculative Decoding with Adaptive Windowing	Speculative decoding accelerates LLM inference by having a lightweight draft model propose speculative windows of candidate tokens for parallel verification by a larger target model. In practice, speculative efficiency is often bottlenecked by hard-to-draft positions, where an early mismatch truncates the accepted pref...	[ "Jie Jiang", "Xing Sun", "Ruotian Chen", "Jianan Su", "Kaixin Shen" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-05-14T00:00:00	https://arxiv.org/abs/2605.14978	https://arxiv.org/pdf/2605.14978v2	2605.14978	null	0	false	null	null	0.35
fd16601b7856f57778cc38080ebd190ab19273c733dd1d25022fcc6626babede	[ "arxiv", "semantic_scholar" ]	Factorization-Error-Free Discrete Diffusion Language Model via Speculative Decoding	Discrete diffusion language models improve generation efficiency through parallel token prediction, but standard $X_0$ prediction methods introduce factorization errors by approximating the clean token posterior with independent token-wise distributions. This paper proposes Factorization-Error-Free Discrete Diffusion L...	[ "Xun Fang", "Yunchen Li", "Hang Yuan", "Zhou Yu" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-05-14T00:00:00	https://arxiv.org/abs/2605.14305	https://arxiv.org/pdf/2605.14305v1	2605.14305	null	0	false	null	null	0.35
9b950f8c90e0109bf8cabf8714f681dce58ea72a67fd29d48a25bb6455820bd5	[ "arxiv", "semantic_scholar" ]	Mistletoe: Stealthy Acceleration-Collapse Attacks on Speculative Decoding	Speculative decoding has become a widely adopted technique for accelerating large language model (LLM) inference by drafting multiple candidate tokens and verifying them with a target model in parallel. Its efficiency, however, critically depends on the average accepted length $τ$, i.e., how many draft tokens survive e...	[ "Shuoyang Sun", "Chang Dai", "Hao Fang", "Kuofeng Gao", "Xinhao Zhong", "Yi Sun", "Fan Mo", "Shu-Tao Xia", "Bin Chen" ]	[ "cs.CL", "cs.LG" ]	[ "Computer Science" ]	2026-05-13T00:00:00	https://arxiv.org/abs/2605.14005	https://arxiv.org/pdf/2605.14005v2	2605.14005	null	0	false	null	null	0.35
567d698b4cee5f5e90c568426a26905573fa25f2e8e073bce18d902f7d66843c	[ "arxiv", "semantic_scholar" ]	PipeSD: An Efficient Cloud-Edge Collaborative Pipeline Inference Framework with Speculative Decoding	Speculative decoding can significantly accelerate LLM inference, especially given that its cloud-edge collaborative deployment offers cloud workload offloading, offline robustness, and privacy enhancement. However, existing collaborative inference frameworks with speculative decoding are constrained by (i) sequential t...	[ "Yunhe Han", "Yunqi Gao", "Bing Hu", "Mahdi Boloursaz Mashhadi", "Yitong Duan", "Pei Xiao", "Yanfeng Zhang" ]	[ "cs.DC" ]	[ "Computer Science" ]	2026-05-13T00:00:00	https://arxiv.org/abs/2605.13319	https://arxiv.org/pdf/2605.13319v3	2605.13319	null	0	false	null	null	0.35
22486e20043601fed8f581f23590b4a874519b61d5542a08d07c0367703c5263	[ "arxiv", "semantic_scholar" ]	CATS: Cascaded Adaptive Tree Speculation for Memory-Limited LLM Inference Acceleration	Auto-regressive decoding in Large Language Models (LLMs) is inherently memory-bound: every generation step requires loading the model weights and intermediate results from memory (e.g., High-Bandwidth Memory (HBM) for GPU servers), making throughput bottlenecked by memory bandwidth rather than compute. Speculative deco...	[ "Yuning Han", "Yangchenchen Jin", "Dylan Zhao", "Jingwei Sun" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-05-11T00:00:00	https://arxiv.org/abs/2605.11186	https://arxiv.org/pdf/2605.11186v1	2605.11186	null	0	false	null	null	0.35
34f0b12c866caa1d6c01d3a9accad01f131880f70807a9fffc022f33b45913de	[ "arxiv", "semantic_scholar" ]	GELATO: Generative Entropy- and Lyapunov-based Adaptive Token Offloading for Device-Edge Speculative LLM Inference	The recent growth of on-device Large Language Model (LLM) inference has driven significant interest in device-edge collaborative LLM inference. As a promising architecture, Speculative Decoding (SD) is increasingly adopted where a lightweight draft model rapidly generates candidate tokens to be verified by a powerful t...	[ "Zengzipeng Tang", "Yuxuan Sun", "Wei Chen", "Jianwen Ding", "Bo Ai" ]	[ "cs.NI", "cs.DC", "cs.IT", "cs.LG" ]	[ "Computer Science", "Mathematics" ]	2026-05-11T00:00:00	https://arxiv.org/abs/2605.10124	https://arxiv.org/pdf/2605.10124v1	2605.10124	null	0	false	null	null	0.35
3ad664e3bced318acf3f428ad5669080dec5a7484cdcf0cd3b181f66a9276b2e	[ "arxiv", "semantic_scholar" ]	Attention Drift: What Autoregressive Speculative Decoding Models Learn	Speculative decoding accelerates LLM inference by drafting future tokens with a small model, but drafter models degrade sharply under template perturbation and long-context inputs. We identify a previously-unreported phenomenon we call \textbf{attention drift}: as the drafter generates successive tokens within a specul...	[ "Doğaç Eldenk", "Payal Mohapatra", "Yigitcan Comlek", "Kaan Oktay", "Hongyang Zhang", "Stephen Xia" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-05-11T00:00:00	https://arxiv.org/abs/2605.09992	https://arxiv.org/pdf/2605.09992v1	2605.09992	null	0	false	null	null	0.35
7d16dde4c1901d12e625e7b87a790624aea83eb9fe3c253ce61dd78f0c4c1b5b	[ "arxiv", "semantic_scholar" ]	SlimSpec: Low-Rank Draft LM-Head for Accelerated Speculative Decoding	Speculative decoding speeds up autoregressive generation in Large Language Models (LLMs) through a two-step procedure, where a lightweight draft model proposes tokens which the target model then verifies in a single forward pass. Although the drafter network is small in modern architectures, its LM-head still performs ...	[ "Anton Plaksin", "Sergei Krutikov", "Sergei Skvortsov", "Alexander Samarin" ]	[ "cs.LG", "cs.CL" ]	[ "Computer Science" ]	2026-05-11T00:00:00	https://arxiv.org/abs/2605.10453	https://arxiv.org/pdf/2605.10453v1	2605.10453	null	0	false	null	null	0.35
dd6635734fc61cb56b5334026e94ff9e387b38ce2d4f3b026ba09b3b91b33d20	[ "arxiv", "semantic_scholar" ]	31.1 A 14.08-to-135.69Token/s ReRAM-on-Logic Stacked Outlier-Free Large-Language-Model Accelerator with Block-Clustered Weight-Compression and Adaptive Parallel-Speculative-Decoding	This work presents a 55nm speculative decoding-based LLM accelerator with bumping-based face-to-face ReRAM-on-logic stacking technology. It features a local rotation unit for outlier-free low-bit quantization, a stacking-aware PNM architecture co-designed with blockwise vector quantization to reduce weight EMA overhead...	[ "Pingcheng Dong", "Yonghao Tan", "Xuejiao Liu", "Peng Luo", "Yu Liu", "Di Pang", "Songchen Ma", "Xijie Huang", "Shih-Yang Liu", "Dong Zhang", "Zhichao Lu", "Luhong Liang", "Chi-Ying Tsui", "Fengbin Tu", "Liang Zhao", "Kwang-Ting Cheng" ]	[ "cs.AR" ]	[ "Computer Science" ]	2026-05-10T00:00:00	https://arxiv.org/abs/2605.09375	https://arxiv.org/pdf/2605.09375v1	2605.09375	10.1109/ISSCC49663.2026.11409211	0	false	null	IEEE International Solid-State Circuits Conference	0.55
a29542d6f8c8ab3c35c60194a6cea3711fde8a615c571236a70ceb258ee53963	[ "arxiv", "semantic_scholar" ]	Test-Time Speculation	Speculative decoding accelerates LLM inference by using a fast draft model to generate tokens and a more accurate target model to verify them. Its performance depends on the $\textit{acceptance length}$, or number of draft tokens accepted by the target. Our studies show that the acceptance length of even state-of-the-a...	[ "Avinash Kumar", "Sujay Sanghavi", "Poulami Das" ]	[ "cs.CL", "cs.LG" ]	[ "Computer Science" ]	2026-05-10T00:00:00	https://arxiv.org/abs/2605.09329	https://arxiv.org/pdf/2605.09329v2	2605.09329	null	0	false	null	null	0.35
5c3811110cc4312de6e3cb80ec3b1ef0fd644caa70282544d91025ce5abbf671	[ "arxiv", "semantic_scholar" ]	PARD-2: Target-Aligned Parallel Draft Model for Dual-Mode Speculative Decoding	Speculative decoding accelerates Large Language Models (LLMs) inference by using a lightweight draft model to propose candidate tokens that are verified in parallel by the target model. However, existing draft model training objectives are not directly aligned with the inference-time goal of maximizing consecutive toke...	[ "Zihao An", "Taichi Liu", "Ziqiong Liu", "Dong Li", "Ruofeng Liu", "Emad Barsoum" ]	[ "cs.CL", "cs.AI" ]	[ "Computer Science" ]	2026-05-09T00:00:00	https://arxiv.org/abs/2605.08632	https://arxiv.org/pdf/2605.08632v1	2605.08632	null	0	true	https://github.com/AMD-AGI/PARD	null	0.65
52c38a8cb1d85c08fb561f42484fd3d3c0e98bc7d34f949d45dfc8b7f51eefb4	[ "arxiv", "semantic_scholar" ]	CASCADE: Context-Aware Relaxation for Speculative Image Decoding	Autoregressive generation is a powerful approach for high-fidelity image synthesis, but it remains computationally demanding and slow even on the most advanced accelerators. While speculative decoding has been explored to mitigate this bottleneck, existing approaches fail to achieve efficiency gains comparable to those...	[ "Selin Yildirim", "Subhajit Dutta Chowdhury", "Mohammad Mahdi Kamani", "Vikram Appia", "Deming Chen" ]	[ "cs.CV", "cs.AI" ]	[ "Computer Science" ]	2026-05-08T00:00:00	https://arxiv.org/abs/2605.07230	https://arxiv.org/pdf/2605.07230v1	2605.07230	null	0	false	null	null	0.35
556e0fb86d22b2ce102091cf9d3573bfcee44200d3faa76548bd3d8e30d348c6	[ "arxiv", "semantic_scholar" ]	Future Validity is the Missing Statistic: From Impossibility to $Φ$-Estimation for Grammar-Faithful Speculative Decoding	Grammar-constrained generation is often combined with local vocabulary masking and speculative decoding, but the resulting sampling law is not the grammar-conditional distribution users usually intend. We show that any speculative decoder with local mask access, Leviathan rejection, and rollback soundness samples from ...	[ "Wenhua Nie", "Zijie Meng", "Kun Zou", "Zheng Lin", "Ziwei Li", "Haoran Zheng", "Jyh-Shing Roger Jang", "Hao Zhang" ]	[ "cs.LG", "cs.IT" ]	[ "Computer Science", "Mathematics" ]	2026-05-08T00:00:00	https://arxiv.org/abs/2605.07698	https://arxiv.org/pdf/2605.07698v1	2605.07698	null	0	false	null	null	0.35
61030784c8e3e6e22215ccee3ac44b6a840d2c81ff8349b6c7d4186a72acc723	[ "arxiv", "semantic_scholar" ]	SpecBlock: Block-Iterative Speculative Decoding with Dynamic Tree Drafting	Speculative decoding accelerates LLM inference by drafting a tree of candidate continuations and verifying it in one target forward. Existing drafters fall into two camps with opposite weaknesses. Autoregressive drafters such as EAGLE-3 preserve dependence along each draft path but call the drafter once per tree depth,...	[ "Weijie Shi", "Qiang Xu", "Fan Deng", "Yaguang Wu", "Jiarun Liu", "Yehong Xu", "Hao Chen", "Jia Zhu", "Jiajie Xu", "Xiangjun Huang", "Jian Yang", "Xiaofang Zhou" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-05-08T00:00:00	https://arxiv.org/abs/2605.07243	https://arxiv.org/pdf/2605.07243v2	2605.07243	null	0	false	null	null	0.35
3a6bdbeafb8b7fb1375eeb609731f435e26de11aca2fc70c7435b2b22499b33d	[ "arxiv", "semantic_scholar" ]	Parallel Prefix Verification for Speculative Generation	We introduce PARSE (PArallel pRefix Speculative Engine), a speculative generation framework that accelerates large language model (LLM) inference by parallelizing prefix verification on a semantic level. Existing speculative decoding methods are fundamentally limited by token-level equivalence: the target model must ve...	[ "Yuncheng Yao", "Yuxuan Xia", "Shengjie Wang", "Danyang Zhuo" ]	[ "cs.AI" ]	[ "Computer Science" ]	2026-05-05T00:00:00	https://arxiv.org/abs/2605.04263	https://arxiv.org/pdf/2605.04263v1	2605.04263	null	0	false	null	null	0.35
d81d6e1d37e9f015a370d0543a05dcc40c1e2b701bf4eb7930a05f67e8c10082	[ "arxiv", "semantic_scholar" ]	SpecKV: Adaptive Speculative Decoding with Compression-Aware Gamma Selection	Speculative decoding accelerates large language model (LLM) inference by using a small draft model to propose candidate tokens that a larger target model verifies. A critical hyperparameter in this process is the speculation length $γ$, which determines how many tokens the draft model proposes per step. Nearly all exis...	[ "Shikhar Shukla" ]	[ "cs.LG", "cs.AI", "cs.CL", "cs.DC", "eess.SY" ]	[ "Computer Science", "Engineering" ]	2026-05-04T00:00:00	https://arxiv.org/abs/2605.02888	https://arxiv.org/pdf/2605.02888v2	2605.02888	null	0	true	https://github.com/Amorfati123/SpecKV	null	0.65
7eb7a5f9140ce7d603710d1b02e2f9077cf214e4dcad0197f32c9d296d186c6c	[ "arxiv", "semantic_scholar" ]	SPECTRE: Hybrid Ordinary-Parallel Speculative Serving for Resource-Efficient LLM Inference	LLM serving platforms are increasingly deployed as multi-model cloud systems, where user demand is often long-tailed: a few popular large models receive most requests, while many smaller tail models remain underutilized. We propose \textbf{SPECTRE} (Parallel \textbf{SPEC}ulative Decoding with a Multi-\textbf{T}enant \t...	[ "Jincheng Xie", "Yawen Ling", "Qi Xiao", "Feiyu Zhang", "Zhongyi Huang", "Wen Hu", "Yu Zheng" ]	[ "cs.DC", "cs.AI" ]	[ "Computer Science" ]	2026-05-04T00:00:00	https://arxiv.org/abs/2605.08151	https://arxiv.org/pdf/2605.08151v2	2605.08151	null	0	true	https://github.com/sgl-project/sglang/pull/22272	null	0.65
f7c2ece51dedc23b887e2321514ae8b6fac663c98779cb2677d30a8ecdda52ed	[ "arxiv", "semantic_scholar" ]	CoVSpec: Efficient Device-Edge Co-Inference for Vision-Language Models via Speculative Decoding	Vision-language models (VLMs) have demonstrated strong capabilities in multimodal perception and reasoning. However, deploying large VLMs on mobile devices remains challenging due to their substantial computational and memory demands. A practical alternative is device-edge co-inference, where a lightweight draft VLM on...	[ "Yuanyuan Jia", "Shunpu Tang", "Qianqian Yang" ]	[ "cs.AI" ]	[ "Computer Science" ]	2026-05-04T00:00:00	https://arxiv.org/abs/2605.02218	https://arxiv.org/pdf/2605.02218v1	2605.02218	null	0	false	null	null	0.35
acc73fea3b2b4266f59431d60739e647587f4f5c55623edc5a55aa7930becbff	[ "arxiv", "semantic_scholar" ]	Component-Aware Self-Speculative Decoding in Hybrid Language Models	Speculative decoding accelerates autoregressive inference by drafting candidate tokens with a fast model and verifying them in parallel with the target. Self-speculative methods avoid the need for an external drafter but have been studied exclusively in homogeneous Transformer architectures. We introduce component-awar...	[ "Hector Borobia", "Elies Seguí-Mas", "Guillermina Tormo-Carbó" ]	[ "cs.CL", "cs.AI" ]	[ "Computer Science" ]	2026-05-01T00:00:00	https://arxiv.org/abs/2605.01106	https://arxiv.org/pdf/2605.01106v1	2605.01106	null	0	true	https://github.com/hecboar/hybrid-speculative-decoding	null	0.65
0b98f82ee509f41751bed8e17e73020891c3538fff332e5bb6b4f462d1f04c0b	[ "arxiv", "semantic_scholar" ]	Making Every Verified Token Count: Adaptive Verification for MoE Speculative Decoding	Tree-based speculative decoding accelerates autoregressive generation by verifying multiple draft candidates in parallel, but this advantage weakens for sparse Mixture-of-Experts (MoE) models. As the draft tree grows, different branches activate different experts, expanding the union of activated experts and substantia...	[ "Lehan Pan", "Ziyang Tao", "Ruoyu Pang", "Xiao Wang", "Jianjun Zhao", "Yanyong Zhang" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-05-01T00:00:00	https://arxiv.org/abs/2605.00342	https://arxiv.org/pdf/2605.00342v1	2605.00342	null	0	false	null	null	0.35
ded1c9f9be141eff511a71ddb73fe8bd9c1aa84402e568afd7a02240e2d62440	[ "arxiv", "semantic_scholar" ]	Accelerating RL Post-Training Rollouts via System-Integrated Speculative Decoding	RL post-training of frontier language models is increasingly bottlenecked by autoregressive rollout generation, making rollout acceleration a central systems challenge. Many existing efficiency methods improve throughput by changing the rollout or optimization regime, for example, through off-policy execution, replay, ...	[ "Hayate Iso", "Tiyasa Mitra", "Sudipta Mondal", "Rasoul Shafipour", "Venmugil Elango", "Terry Kong", "Yuki Huang", "Seonjin Na", "Izzy Putterman", "Benjamin Chislett", "Maor Ashkenazi", "Joseph Guman", "Gerald Shen", "Tugrul Konuk", "Ashwath Aithal", "Ritika Borkar", "Ran Zilberstein...	[ "cs.LG", "cs.CL" ]	[ "Computer Science" ]	2026-04-29T00:00:00	https://arxiv.org/abs/2604.26779	https://arxiv.org/pdf/2604.26779v1	2604.26779	10.48550/arXiv.2604.26779	2	false	null	arXiv.org	0.55
460a2479e928c9b61782fb98c048b4f071996aab29820ffca4d2d6313b8e4995	[ "arxiv", "semantic_scholar" ]	When Hidden States Drift: Can KV Caches Rescue Long-Range Speculative Decoding?	Speculative decoding accelerates LLM inference, but SOTA hidden-state-based drafters suffer from long-range decay: draft accuracy degrades as the speculative step increases. Existing work attributes this decay to train-inference mismatch and proposes test-time training (TTT) as a remedy, yet we observe that long-range ...	[ "Tianyu Liu", "Yuhao Shen", "Xinyi Hu", "Baolin Zhang", "Hengxin Zhang", "Jun Dai", "Jun Zhang", "Shuang Ge", "Lei Chen", "Yue Li", "MingCheng Wan" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-04-29T00:00:00	https://arxiv.org/abs/2604.26412	https://arxiv.org/pdf/2604.26412v2	2604.26412	10.48550/arXiv.2604.26412	1	false	null	arXiv.org	0.55
3444375581815c62ca166972a5aeb5e3b90d11b34a941f4804cf9e1815bfe138	[ "arxiv", "semantic_scholar" ]	An Empirical Study of Speculative Decoding on Software Engineering Tasks	Large Language Models (LLMs) have become widely used for Software Engineering (SE) tasks, spanning from function-level code generation to complex repository-level workflows. However, the high latency of autoregressive inference remains a significant bottleneck, hindering their deployment in interactive environments. Wh...	[ "Yijia Li", "Junkai Chen", "Xing Hu", "Xin Xia" ]	[ "cs.SE" ]	[ "Computer Science" ]	2026-04-29T00:00:00	https://arxiv.org/abs/2604.26469	https://arxiv.org/pdf/2604.26469v3	2604.26469	10.48550/arXiv.2604.26469	0	false	null	arXiv.org	0.55
b762005c1bb065e44f1f3b490d7b31badc0882f1c0e7a1ccf1733c659b6b3b45	[ "arxiv", "semantic_scholar" ]	SpecFed: Accelerating Federated LLM Inference with Speculative Decoding and Compressed Transmission	Federated inference enhances LLM performance in edge computing through weighted averaging of distributed model predictions. However, autoregressive LLM inference requires frequent full-model forward passes across workers, severely limiting decoding throughput. Distributed deployment further aggravates this due to a com...	[ "Ce Zheng", "Xinghan Wang", "Jiahong Ning", "Yuxuan Shi", "Ning Huang", "Tingting Yang" ]	[ "eess.SP", "cs.DC" ]	[ "Computer Science", "Engineering" ]	2026-04-28T00:00:00	https://arxiv.org/abs/2604.25777	https://arxiv.org/pdf/2604.25777v1	2604.25777	10.48550/arXiv.2604.25777	0	false	null	arXiv.org	0.55
e43cf4d320579220f4880a28fbaa065ddf4dd98b3d1a323dcdcc719ba68cd9e4	[ "arxiv", "semantic_scholar" ]	AHASD: Asynchronous Heterogeneous Architecture for LLM Adaptive Drafting Speculative Decoding on Mobile Devices	Speculative decoding enhances the inference efficiency of large language models (LLMs) by generating drafts using a small draft language model (DLM) and verifying them in batches with a large target language model (TLM). However, adaptive drafting inference on a mobile single-NPU-PIM system faces idle overhead in tradi...	[ "Ma Zirui", "Fan Zhihua", "Li Wenxing", "Wu Haibin", "Zhang Fulin", "Ye Xiaochun", "Li Wenming" ]	[ "cs.AR", "cs.AI" ]	[ "Computer Science" ]	2026-04-28T00:00:00	https://arxiv.org/abs/2604.25326	https://arxiv.org/pdf/2604.25326v3	2604.25326	10.1145/3770743.3803965	0	true	https://github.com/MAdrid1011/AHASD	arXiv.org	0.85
a809065401a8ed4e20d45eb999e222ac3577319fa1245a226ef27cd9450c2b88	[ "arxiv", "semantic_scholar" ]	FASER: Fine-Grained Phase Management for Speculative Decoding in Dynamic LLM Serving	Speculative decoding (SD) is a widely used approach for accelerating decode-heavy LLM inference workloads. While online inference workloads are highly dynamic, existing SD systems are rigid and take a coarse-grained approach to SD management. They typically set the speculative token length for an entire batch and seria...	[ "Wenyan Chen", "Chengzhi Lu", "Yanying Lin", "Dmitrii Ustiugov" ]	[ "cs.DC" ]	[ "Computer Science" ]	2026-04-22T00:00:00	https://arxiv.org/abs/2604.20503	https://arxiv.org/pdf/2604.20503v1	2604.20503	10.48550/arXiv.2604.20503	0	false	null	arXiv.org	0.55
dd1acd20e7b27782a72bf65ecaa7f0f3446db53948d30fe37eba90bfa0f505e7	[ "arxiv", "semantic_scholar" ]	DiP-SD: Distributed Pipelined Speculative Decoding for Efficient LLM Inference at the Edge	Speculative decoding has emerged as a promising technique for large language model (LLM) inference by accelerating autoregressive decoding via draft-then-verify. This paper studies a new edge scenario with multi-user inference, where draft tokens are generated locally on devices and subsequently offloaded to a centrali...	[ "Yaodan Xu", "Sheng Zhou", "Zhisheng Niu" ]	[ "cs.IT" ]	[ "Computer Science", "Mathematics" ]	2026-04-22T00:00:00	https://arxiv.org/abs/2604.20919	https://arxiv.org/pdf/2604.20919v1	2604.20919	10.48550/arXiv.2604.20919	0	false	null	arXiv.org	0.55
7449dd94dbbb12be98fd6837c186930f2a831cd1a1aa91815547cff5fea264ee	[ "arxiv", "semantic_scholar" ]	LLM-Viterbi: Semantic-Aware Decoding for Convolutional Codes	Traditional wireless communications rely solely on bit-level channel coding for error correction, without exploiting the inherent linguistic structure of the data source. This paper proposes a large language model (LLM) Viterbi decoder that integrates LLM priors into the Viterbi decoding for text transmission over AWGN...	[ "Zhengtong Li", "Chentao Yue", "Jiafu Hao", "Branka Vucetic", "Yonghui Li" ]	[ "cs.IT" ]	[ "Computer Science", "Mathematics" ]	2026-04-21T00:00:00	https://arxiv.org/abs/2604.19035	https://arxiv.org/pdf/2604.19035v1	2604.19035	10.48550/arXiv.2604.19035	1	true	https://github.com/Todd-6/LLM-Viterbi	arXiv.org	0.85
eff37c71bb5300f9d772a02406e08c03f6ddf26ef24cb0c7def66604a9277e81	[ "arxiv", "semantic_scholar" ]	WISV: Wireless-Informed Semantic Verification for Distributed Speculative Decoding in Device-Edge LLM Inference	While distributed device-edge speculative decoding enhances resource utilization across heterogeneous nodes, its performance is often bottlenecked by conventional token-level verification strategies. Such rigid alignment leads to excessive rejections, significantly diminishing the accepted sequence length and increasin...	[ "Zixuan Liu", "Zhiyong Chen", "Nan Xue", "Shengkang Chen", "Jiangchao Yao", "Meixia Tao", "Wenjun Zhang" ]	[ "cs.IT", "cs.AI" ]	[ "Computer Science", "Mathematics" ]	2026-04-20T00:00:00	https://arxiv.org/abs/2604.17701	https://arxiv.org/pdf/2604.17701v1	2604.17701	10.48550/arXiv.2604.17701	1	false	null	arXiv.org	0.55
50311ae68c1790c8d319f811a79a0560cbf138c1c1215fd7696c6e1039c13080	[ "arxiv", "semantic_scholar" ]	Speculative Decoding for Autoregressive Video Generation	Autoregressive video diffusion is emerging as a promising paradigm for streaming video synthesis, with step distillation serving as the primary means of accelerating inference. Whether speculative decoding, the dominant acceleration strategy for large language models, can be effectively adapted to autoregressive video ...	[ "Yuezhou Hu", "Jintao Zhang" ]	[ "cs.CV", "cs.AI" ]	[ "Computer Science" ]	2026-04-19T00:00:00	https://arxiv.org/abs/2604.17397	https://arxiv.org/pdf/2604.17397v1	2604.17397	10.48550/arXiv.2604.17397	0	false	null	arXiv.org	0.5489
f72a41800ee2e901b6e9ea04de7ec99806dd37b891352b6e55e02772b9261cc8	[ "arxiv", "semantic_scholar" ]	Faster LLM Inference via Sequential Monte Carlo	Speculative decoding (SD) accelerates language model inference by drafting tokens from a cheap proposal model and verifying them against an expensive target model via rejection sampling. Because rejection truncates the draft block at the first error, throughput degrades when draft and target diverge. Rather than reject...	[ "Yahya Emara", "Mauricio Barba da Costa", "Chi-Chih Chang", "Cameron Freer", "Tim Vieira", "Ryan Cotterell", "Mohamed S. Abdelfattah" ]	[ "cs.LG", "cs.CL" ]	[ "Computer Science" ]	2026-04-17T00:00:00	https://arxiv.org/abs/2604.15672	https://arxiv.org/pdf/2604.15672v1	2604.15672	10.48550/arXiv.2604.15672	0	false	null	arXiv.org	0.5466
11c4239976d3da1d0ae54c133e70c465a42e83efccb1c40bef43d4595c3a07d9	[ "arxiv", "semantic_scholar" ]	EvoSpec: Evolving Speculative Decoding via Real-Time Vocabulary and Parameter Adaptation	Speculative decoding accelerates Large Language Model inference via a draft-then-verify paradigm, yet the output projection layer becomes a bottleneck as vocabulary sizes scale. While existing static pruning methods effectively reduce this overhead, they suffer from precipitous drops in acceptance rate in specialized d...	[ "Shuyu Zhang", "Lingfeng Pan", "Qicheng Wang", "Yaqi Shi", "Yueyang Tan", "Ruyu Yan", "Jiaqi Chen", "Lixing Du", "Lu Wang" ]	[ "cs.CL", "cs.AI" ]	[ "Computer Science" ]	2026-04-17T00:00:00	https://arxiv.org/abs/2605.27390	https://arxiv.org/pdf/2605.27390v2	2605.27390	null	0	false	null	null	0.3478
96a8ba45ae2a2e9e3b1e291b0ddb98ee2018cafd2dc910bbb6c3d851d8251b71	[ "arxiv", "semantic_scholar" ]	RACER: Retrieval-Augmented Contextual Rapid Speculative Decoding	Autoregressive decoding in Large Language Models (LLMs) generates one token per step, causing high inference latency. Speculative decoding (SD) mitigates this through a guess-and-verify strategy, but existing training-free variants face trade-offs: retrieval-based drafts break when no exact match exists, while logits-b...	[ "Zihong Zhang", "Zuchao Li", "Lefei Zhang", "Ping Wang", "Hai Zhao" ]	[ "cs.CL", "cs.AI" ]	[ "Computer Science" ]	2026-04-16T00:00:00	https://arxiv.org/abs/2604.14885	https://arxiv.org/pdf/2604.14885v1	2604.14885	10.48550/arXiv.2604.14885	1	true	https://github.com/hkr04/RACER}{https://github.com/hkr04/RACER}$	arXiv.org	0.8429
ff11338d64b751e1e89defdd12f992c998af0a5a52ca22a5f8f535b99839070d	[ "arxiv", "semantic_scholar" ]	Acceptance Dynamics Across Cognitive Domains in Speculative Decoding	Speculative decoding accelerates large language model (LLM) inference. It uses a small draft model to propose a tree of future tokens. A larger target model then verifies these tokens in a single batched forward pass. Despite the growing body of work on speculative methods, the degree to which the cognitive characteris...	[ "Saif Mahmoud" ]	[ "cs.AI", "cs.CL" ]	[ "Computer Science" ]	2026-04-16T00:00:00	https://arxiv.org/abs/2604.14682	https://arxiv.org/pdf/2604.14682v1	2604.14682	10.48550/arXiv.2604.14682	0	false	null	arXiv.org	0.5454
ea1fd3b3a361eec707cee3fbad1b5fe03b0a7edff2022ad3e8586671e11dd7b9	[ "arxiv", "semantic_scholar" ]	From Tokens to Steps: Verification-Aware Speculative Decoding for Efficient Multi-Step Reasoning	Speculative decoding (SD) accelerates large language model inference by allowing a lightweight draft model to propose outputs that a stronger target model verifies. However, its token-centric nature allows erroneous steps to propagate. Prior approaches mitigate this using external reward models, but incur additional la...	[ "Kiran Purohit", "Ramasuri Narayanam", "Soumyabrata Pal" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-04-16T00:00:00	https://arxiv.org/abs/2604.15244	https://arxiv.org/pdf/2604.15244v1	2604.15244	10.48550/arXiv.2604.15244	0	false	null	arXiv.org	0.5454
71f962a2fbfbf4fae4f1964e1f136c1654e986a8b5437e30bd606d975f68e17d	[ "arxiv", "semantic_scholar" ]	ConfLayers: Adaptive Confidence-based Layer Skipping for Self-Speculative Decoding	Self-speculative decoding is an inference technique for large language models designed to speed up generation without sacrificing output quality. It combines fast, approximate decoding using a compact version of the model as a draft model with selective re-evaluation by the full target model. Some existing methods form...	[ "Walaa Amer", "Uday das", "Fadi Kurdahi" ]	[ "cs.LG", "cs.CL" ]	[ "Computer Science" ]	2026-04-16T00:00:00	https://arxiv.org/abs/2604.14612	https://arxiv.org/pdf/2604.14612v1	2604.14612	10.48550/arXiv.2604.14612	2	false	null	arXiv.org	0.5454
e24d78808fe21c52724f192798875a78c89a5d2a9e29f75d5c7214ad3235ec60	[ "arxiv", "semantic_scholar" ]	ELMoE-3D: Leveraging Intrinsic Elasticity of MoE for Hybrid-Bonding-Enabled Self-Speculative Decoding in On-Premises Serving	Mixture-of-Experts (MoE) models have become the dominant architecture for large-scale language models, yet on-premises serving remains fundamentally memory-bound as batching turns sparse per-token compute into dense memory activation. Memory-centric architectures (PIM, NMP) improve bandwidth but leave compute underutil...	[ "Yuseon Choi", "Jingu Lee", "Jungjun Oh", "Sunjoo Whang", "Byeongcheol Kim", "Minsung Kim", "Hoi-Jun Yoo", "Sangjin Kim" ]	[ "cs.LG", "cs.AI", "cs.AR", "cs.DC" ]	[ "Computer Science" ]	2026-04-16T00:00:00	https://arxiv.org/abs/2604.14626	https://arxiv.org/pdf/2604.14626v2	2604.14626	10.48550/arXiv.2604.14626	0	false	null	arXiv.org	0.5454
5cca20b3d9d46eca9c699f6df570933331a5dd66f22de603d0a045fc28497734	[ "arxiv", "semantic_scholar" ]	Calibrated Speculative Decoding: Frequency-Guided Candidate Selection for Efficient Inference	Speculative decoding accelerates autoregressive generation by letting draft tokens bypass full verification, but conventional frameworks suffer from frequent false rejections, particularly when draft models produce semantically correct but lexically divergent outputs. In this paper, we present Calibrated Speculative De...	[ "Xuwen Zhou", "Fangxin Liu", "Chao Wang", "Xiao Zheng", "Hao Zheng", "Min He", "Li Jiang", "Haibing Guan" ]	[ "cs.CL", "cs.LG" ]	[ "Computer Science" ]	2026-04-15T00:00:00	https://arxiv.org/abs/2604.13634	https://arxiv.org/pdf/2604.13634v1	2604.13634	10.48550/arXiv.2604.13634	0	false	null	arXiv.org	0.5443
2f923df658b8fb3996804d613961ca293ca8d83b52e4944244dd8beed1c66cab	[ "arxiv", "semantic_scholar" ]	ToolSpec: Accelerating Tool Calling via Schema-Aware and Retrieval-Augmented Speculative Decoding	Tool calling has greatly expanded the practical utility of large language models (LLMs) by enabling them to interact with external applications. As LLM capabilities advance, effective tool use increasingly involves multi-step, multi-turn interactions to solve complex tasks. However, the resulting growth in tool interac...	[ "Heming Xia", "Yongqi Li", "Cunxiao Du", "Mingbo Song", "Wenjie Li" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-04-15T00:00:00	https://arxiv.org/abs/2604.13519	https://arxiv.org/pdf/2604.13519v2	2604.13519	10.48550/arXiv.2604.13519	1	false	null	arXiv.org	0.5443
8c95dc8000e2c620d3bdf43001fdcbdcacae537b921dc28c99367e816acf72d9	[ "arxiv", "semantic_scholar" ]	Accelerating Speculative Decoding with Block Diffusion Draft Trees	Speculative decoding accelerates autoregressive language models by using a lightweight drafter to propose multiple future tokens, which the target model then verifies in parallel. DFlash shows that a block diffusion drafter can generate an entire draft block in a single forward pass and achieve state-of-the-art specula...	[ "Liran Ringel", "Yaniv Romano" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-04-14T00:00:00	https://arxiv.org/abs/2604.12989	https://arxiv.org/pdf/2604.12989v1	2604.12989	10.48550/arXiv.2604.12989	4	false	null	arXiv.org	0.5431
c029f2bf77b0863e78c5a12a5c361abf01587f2cc84cb92e9ef4b3bb3336dfcb	[ "arxiv", "semantic_scholar" ]	SENSE: Semantic Embedding Navigation with Soft-gated Evaluation for Retrieval-based Speculative Decoding	Speculative Decoding (SD) accelerates Large Language Model (LLM) inference by employing a lightweight draft model to propose candidate tokens, which are verified in parallel by the target model, without compromising generation quality. While Retrieval-based Speculative Decoding (RSD) is favored for its plug-and-play ve...	[ "Shaowen Chen", "Zhicheng Liao", "Hongwei Wang" ]	[ "cs.CL", "cs.AI", "cs.LG" ]	[ "Computer Science" ]	2026-04-14T00:00:00	https://arxiv.org/abs/2606.00021	https://arxiv.org/pdf/2606.00021v1	2606.00021	null	0	false	null	null	0.3456
e4b7bad1d0e5c499d78182cd4e54e27ee1d87b9c954936d69b2bb8bc3fb76c74	[ "arxiv", "semantic_scholar" ]	SOLARIS: Speculative Offloading of Latent-bAsed Representation for Inference Scaling	Recent advances in recommendation scaling laws have led to foundation models of unprecedented complexity. While these models offer superior performance, their computational demands make real-time serving impractical, often forcing practitioners to rely on knowledge distillation-compromising serving quality for efficien...	[ "Zikun Liu", "Liang Luo", "Qianru Li", "Zhengyu Zhang", "Wei Ling", "Jingyi Shen", "Zeliang Chen", "Yaning Huang", "Jingxian Huang", "Abdallah Aboelela", "Chonglin Sun", "Feifan Gu", "Fenggang Wu", "Hang Qu", "Huayu Li", "Jill Pan", "Kaidi Pei", "Laming Chen", "Longhao Jin", "Q...	[ "cs.LG" ]	[ "Computer Science" ]	2026-04-13T00:00:00	https://arxiv.org/abs/2604.12110	https://arxiv.org/pdf/2604.12110v2	2604.12110	10.48550/arXiv.2604.12110	0	false	null	arXiv.org	0.542
43a41d9db3e337a905ba1e7c42a98b0ac7f64cb56ff4b3c93b96769b15eaae26	[ "arxiv", "semantic_scholar" ]	SpecMoE: A Fast and Efficient Mixture-of-Experts Inference via Self-Assisted Speculative Decoding	The Mixture-of-Experts (MoE) architecture has emerged as a promising approach to mitigate the rising computational costs of large language models (LLMs) by selectively activating parameters. However, its high memory requirements and sub-optimal parameter efficiency pose significant challenges for efficient deployment. ...	[ "Jehyeon Bang", "Eunyeong Cho", "Ranggi Hwang", "Jinha Chung", "Minsoo Rhu" ]	[ "cs.AI", "cs.LG" ]	[ "Computer Science" ]	2026-04-11T00:00:00	https://arxiv.org/abs/2604.10152	https://arxiv.org/pdf/2604.10152v1	2604.10152	10.48550/arXiv.2604.10152	0	false	null	arXiv.org	0.5397
4d16de4c79e9bca2245057aa3a10cb125e3eae867fe128fc727acdb23325f531	[ "arxiv", "semantic_scholar" ]	DIVERSED: Relaxed Speculative Decoding via Dynamic Ensemble Verification	Speculative decoding is an effective technique for accelerating large language model inference by drafting multiple tokens in parallel. In practice, its speedup is often bottlenecked by a rigid verification step that strictly enforces the accepted token distribution to exactly match the target model. This constraint le...	[ "Ziyi Wang", "Siva Rajesh Kasa", "Ankith M S", "Santhosh Kumar Kasa", "Jiaru Zou", "Sumit Negi", "Ruqi Zhang", "Nan Jiang", "Qifan Song" ]	[ "cs.CL", "cs.AI", "cs.LG" ]	[ "Computer Science" ]	2026-04-08T00:00:00	https://arxiv.org/abs/2604.07622	https://arxiv.org/pdf/2604.07622v1	2604.07622	10.48550/arXiv.2604.07622	1	true	https://github.com/comeusr/diversed	arXiv.org	0.8287
b317d3ab008f1b442542c4d81d4c7b801ecd768774a3840a61cf1f57bebe987f	[ "arxiv", "semantic_scholar" ]	ConfigSpec: Profiling-Based Configuration Selection for Distributed Edge--Cloud Speculative LLM Serving	Speculative decoding enables collaborative Large Language Model (LLM) inference across cloud and edge by separating lightweight token drafting from heavyweight verification. While prior systems show performance and cost benefits, practical deployment requires navigating a large configuration space spanning draft model ...	[ "Xiangchen Li", "Saeid Ghafouri", "Jiakun Fan", "Babar Ali", "Hans Vandierendonck", "Dimitrios S. Nikolopoulos" ]	[ "cs.DC", "cs.AI" ]	[ "Computer Science" ]	2026-04-08T00:00:00	https://arxiv.org/abs/2604.09722	https://arxiv.org/pdf/2604.09722v1	2604.09722	10.1145/3802513.3803483	0	false	null	null	0.3412
91d22f655e9b6cffc3cd2efe44d13a8207676a491cb521b5467a0644748e4e28	[ "arxiv", "semantic_scholar" ]	NanoSpec: Accelerating Speculative Decoding using Minimalist In-Context Vocabularies	The massive vocabulary sizes of large language models, often exceeding 100k tokens, impose a computational bottleneck on the final linear projection layer during speculative decoding. Existing vocabulary pruning solutions rely on static or coarsely-grained sub-vocabularies that necessitate large active sizes ($\sim$30k...	[ "Zhiyang Chen", "Daliang Xu", "Yinyuan Zhang", "Chenghua Wang", "Mengwei Xu", "Yun Ma" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-04-08T00:00:00	https://arxiv.org/abs/2605.26444	https://arxiv.org/pdf/2605.26444v2	2605.26444	null	0	false	null	null	0.3412
c873da0c66f4c88ca9b2de481640ff59ea34cdd6ef0cee77400ba147da820047	[ "arxiv", "semantic_scholar" ]	See the Forest for the Trees: Loosely Speculative Decoding via Visual-Semantic Guidance for Efficient Inference of Video LLMs	Video Large Language Models (Video-LLMs) excel in video understanding but suffer from high inference latency during autoregressive generation. Speculative Decoding (SD) mitigates this by applying a draft-and-verify paradigm, yet existing methods are constrained by rigid exact-match rules, severely limiting the accelera...	[ "Yicheng Ji", "Jun Zhang", "Jinpeng Chen", "Cong Wang", "Lidan Shou", "Gang Chen", "Huan Li" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-04-07T00:00:00	https://arxiv.org/abs/2604.05650	https://arxiv.org/pdf/2604.05650v2	2604.05650	10.48550/arXiv.2604.05650	3	false	null	arXiv.org	0.5351
f7654b7effe7ed608afc581723dc36692465f2d861631e2f440f5a81f87e5652	[ "arxiv", "semantic_scholar" ]	Multi-Drafter Speculative Decoding with Alignment Feedback	Speculative decoding (SD) accelerates large language model (LLM) inference by using a smaller model to draft future tokens, which are then verified by the target LLM. This preserves generation quality by accepting only aligned tokens. However, individual drafters, often trained for specific tasks or domains, exhibit li...	[ "Taehyeon Kim", "Hojung Jung", "Se-Young Yun" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-04-07T00:00:00	https://arxiv.org/abs/2604.05417	https://arxiv.org/pdf/2604.05417v1	2604.05417	10.48550/arXiv.2604.05417	1	false	null	arXiv.org	0.5351
853edba5b9debb95b784b406e5131475cede78b0682ac41c8344994f6e9e8441	[ "arxiv", "semantic_scholar" ]	Epistemic Blinding: An Inference-Time Protocol for Auditing Prior Contamination in LLM-Assisted Analysis	This paper presents epistemic blinding in the context of an agentic system that uses large language models to reason across multiple biological datasets for drug target prioritization. During development, it became apparent that LLM outputs silently blend data-driven inference with memorized priors about named entities...	[ "Michael Cuccarese" ]	[ "cs.AI", "cs.CL" ]	[ "Computer Science" ]	2026-04-07T00:00:00	https://arxiv.org/abs/2604.06013	https://arxiv.org/pdf/2604.06013v1	2604.06013	10.48550/arXiv.2604.06013	0	true	https://github.com/mcuccarese/epistemic-blinding	arXiv.org	0.827
67c2514fad7827b15b9bccca9d88b97e3b9525d755d92f93b35fc2b0ee19ca7e	[ "arxiv", "semantic_scholar" ]	DualDiffusion: A Speculative Decoding Strategy for Masked Diffusion Models	Masked Diffusion Models (MDMs) offer a promising alternative to autoregressive language models by enabling parallel token generation and bidirectional context modeling. However, their inference speed is significantly limited by the inability to cache key-value pairs due to bidirectional attention, requiring $O(N^2)$ co...	[ "Satyam Goyal", "Kushal Patel", "Tanush Mittal", "Arjun Laxman" ]	[ "cs.LG", "cs.CL" ]	[ "Computer Science" ]	2026-04-06T00:00:00	https://arxiv.org/abs/2604.05250	https://arxiv.org/pdf/2604.05250v1	2604.05250	10.48550/arXiv.2604.05250	0	false	null	arXiv.org	0.534
30ce6022ff2dfe54ed50bcd76bf29f72f46be2f9ed6649bc9c49fd1eca86369c	[ "arxiv", "semantic_scholar" ]	Cactus: Accelerating Auto-Regressive Decoding with Constrained Acceptance Speculative Sampling	Speculative sampling (SpS) has been successful in accelerating the decoding throughput of auto-regressive large language models by leveraging smaller draft models. SpS strictly enforces the generated distribution to match that of the verifier LLM. This is unnecessarily restrictive as slight variations of the verifier's...	[ "Yongchang Hao", "Lili Mou" ]	[ "cs.LG", "cs.AI", "math.OC", "stat.ML" ]	[ "Computer Science", "Mathematics" ]	2026-04-05T00:00:00	https://arxiv.org/abs/2604.04987	https://arxiv.org/pdf/2604.04987v1	2604.04987	10.48550/arXiv.2604.04987	0	false	null	arXiv.org	0.5328
c027a2df4835c9936edcc72c0cab4ae6defe2d2d19e6e70104b581c1777b6b27	[ "arxiv", "semantic_scholar" ]	Goose: Anisotropic Speculation Trees for Training-Free Speculative Decoding	Speculative decoding accelerates large language model inference by drafting multiple candidate tokens and verifying them in a single forward pass. Candidates are organized as a tree: deeper trees accept more tokens per step, but adding depth requires sacrificing breadth (fallback options) under a fixed verification bud...	[ "Tao Jin", "Phuong Minh Nguyen", "Naoya Inoue" ]	[ "cs.CL", "cs.AI" ]	[ "Computer Science" ]	2026-04-02T00:00:00	https://arxiv.org/abs/2604.02047	https://arxiv.org/pdf/2604.02047v1	2604.02047	10.48550/arXiv.2604.02047	1	false	null	arXiv.org	0.5294
3a497fbfb0502bfd45305ceb09c1bb9ade58a190561a068d502c828d4f7769c6	[ "arxiv", "semantic_scholar" ]	SpecTr-GBV: Multi-Draft Block Verification Accelerating Speculative Decoding	Autoregressive language models suffer from high inference latency due to their sequential decoding nature. Speculative decoding (SD) mitigates this by employing a lightweight draft model to propose candidate tokens, which are selectively verified by a larger target model. While existing methods either adopt multi-draft...	[ "Yijun Lin", "Jinhao Sheng", "Qingyue Cai", "Feng Zhou" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-04-01T00:00:00	https://arxiv.org/abs/2604.25925	https://arxiv.org/pdf/2604.25925v1	2604.25925	10.48550/arXiv.2604.25925	0	false	null	arXiv.org	0.5282
3d752b08d43b132c7377162cb1892a015776414478211ccb8623a86c45dd53c0	[ "arxiv", "semantic_scholar" ]	SJD-VP: Speculative Jacobi Decoding with Verification Prediction for Autoregressive Image Generation	Speculative Jacobi Decoding (SJD) has emerged as a promising method for accelerating autoregressive image generation. Despite its potential, existing SJD approaches often suffer from the low acceptance rate issue of speculative tokens due to token selection ambiguity. Recent works attempt to mitigate this issue primari...	[ "Bingqi Shan", "Baoquan Zhang", "Xiaochen Qi", "Xutao Li", "Yunming Ye", "Liqiang Nie" ]	[ "cs.CV" ]	[ "Computer Science" ]	2026-03-28T00:00:00	https://arxiv.org/abs/2603.27115	https://arxiv.org/pdf/2603.27115v1	2603.27115	10.48550/arXiv.2603.27115	0	false	null	arXiv.org	0.5236
c250d01b0d9a40fa17b9da9abc452c172dc733f062274fabe65ab63a70952883	[ "arxiv", "semantic_scholar" ]	Accelerating PayPal's Commerce Agent with Speculative Decoding: An Empirical Study on EAGLE3 with Fine-Tuned Nemotron Models	We evaluate speculative decoding with EAGLE3 as an inference-time optimization for PayPal's Commerce Agent, powered by a fine-tuned llama3.1-nemotron-nano-8B-v1 model. Building on prior work (NEMO-4-PAYPAL) that reduced latency and cost through domain-specific fine-tuning, we benchmark EAGLE3 via vLLM against NVIDIA NI...	[ "Ally Qin", "Jian Wan", "Sarat Mudunuri", "Srinivasan Manoharan" ]	[ "cs.LG", "cs.AI" ]	[ "Computer Science" ]	2026-03-27T00:00:00	https://arxiv.org/abs/2604.19767	https://arxiv.org/pdf/2604.19767v1	2604.19767	10.48550/arXiv.2604.19767	0	false	null	arXiv.org	0.5225
bdae8340baa21b72b415b7a3551242849ab776db4818aab69c35557611755ea4	[ "arxiv", "semantic_scholar" ]	S2D2: Fast Decoding for Diffusion LLMs via Training-Free Self-Speculation	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: aggressi...	[ "Ligong Han", "Hao Wang", "Han Gao", "Kai Xu", "Akash Srivastava" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-03-26T00:00:00	https://arxiv.org/abs/2603.25702	https://arxiv.org/pdf/2603.25702v2	2603.25702	10.48550/arXiv.2603.25702	1	true	https://github.com/phymhan/S2D2	arXiv.org	0.8057
4a38df3d05a310265387354c0143f000c1a40a3d03e88171e55ac7efca89fa73	[ "arxiv", "semantic_scholar" ]	Cross-Family Speculative Decoding for Polish Language Models on Apple~Silicon: An Empirical Evaluation of Bielik~11B with UAG-Extended MLX-LM	Speculative decoding accelerates LLM inference by using a small draft model to propose k candidate tokens for a target model to verify. While effective for same-tokenizer pairs on high-bandwidth GPUs, its applicability to cross-family pairs with mismatched tokenizers and consumer-grade unified memory remains underexplo...	[ "Krzysztof Fonal" ]	[ "cs.CL" ]	[ "Computer Science" ]	2026-03-22T00:00:00	https://arxiv.org/abs/2604.16368	https://arxiv.org/pdf/2604.16368v2	2604.16368	10.48550/arXiv.2604.16368	0	false	null	arXiv.org	0.5168
cfe21e34399c12f1675a9430a3aa83aa511cf579cd960b145f080e20e98065ae	[ "arxiv", "semantic_scholar" ]	ParallelVLM: Lossless Video-LLM Acceleration with Visual Alignment Aware Parallel Speculative Decoding	Although current Video-LLMs achieve impressive performance in video understanding tasks, their autoregressive decoding efficiency remains constrained by the massive number of video tokens. Visual token pruning can partially ease this bottleneck, yet existing approaches still suffer from information loss and yield only ...	[ "Quan Kong", "Yuhao Shen", "Yicheng Ji", "Huan Li", "Cong Wang" ]	[ "cs.CV" ]	[ "Computer Science" ]	2026-03-20T00:00:00	https://arxiv.org/abs/2603.19610	https://arxiv.org/pdf/2603.19610v2	2603.19610	10.48550/arXiv.2603.19610	5	false	null	arXiv.org	0.5145
9ec11f88936ceb5debb2fc82a2b85a9a258463900f58df8d8380f032b8400920	[ "arxiv", "semantic_scholar" ]	A Pipelined Collaborative Speculative Decoding Framework for Efficient Edge-Cloud LLM Inference	Recent advancements and widespread adoption of Large Language Models (LLMs) in both industry and academia have catalyzed significant demand for LLM serving. However, traditional cloud services incur high costs, while on-device inference alone faces challenges due to limited resources. Edge-cloud collaboration emerges a...	[ "Yida Zhang", "Zhiyong Gao", "Shuaibing Yue", "Jie Li", "Rui Wang" ]	[ "cs.DC" ]	[ "Computer Science" ]	2026-03-19T00:00:00	https://arxiv.org/abs/2603.19133	https://arxiv.org/pdf/2603.19133v2	2603.19133	10.48550/arXiv.2603.19133	0	false	null	arXiv.org	0.5133
f41fea8a25a5f7a5bade1a111f0c3b7c5a231af4297bbd0884723459ec875e04	[ "arxiv", "semantic_scholar" ]	SpecForge: A Flexible and Efficient Open-Source Training Framework for Speculative Decoding	Large language models incur high inference latency due to sequential autoregressive decoding. Speculative decoding alleviates this bottleneck by using a lightweight draft model to propose multiple tokens for batched verification. However, its adoption has been limited by the lack of high-quality draft models and scalab...	[ "Shenggui Li", "Chao Wang", "Yikai Zhu", "Yubo Wang", "Fan Yin", "Shuai Shi", "Yefei Chen", "Xiaomin Dong", "Qiaoling Chen", "Jin Pan", "Ji Li", "Laixin Xie", "Yineng Zhang", "Lei Yu", "Yonggang Wen", "Ivor Tsang", "Tianwei Zhang" ]	[ "cs.LG", "cs.AI", "cs.CL" ]	[ "Computer Science" ]	2026-03-19T00:00:00	https://arxiv.org/abs/2603.18567	https://arxiv.org/pdf/2603.18567v1	2603.18567	10.48550/arXiv.2603.18567	4	true	null	arXiv.org	0.7933
88d6a550b2363922bfcd189efc53766d327fd7a9668ef999cd090974fd5a9516	[ "arxiv", "semantic_scholar" ]	MMSpec: Benchmarking Speculative Decoding for Vision-Language Models	Vision-language models (VLMs) achieve strong performance on multimodal tasks but suffer from high inference latency due to large model sizes and long multimodal contexts. Speculative decoding has recently emerged as an effective acceleration technique, yet its behavior in VLMs remains insufficiently understood. We intr...	[ "Hui Shen", "Xin Wang", "Ping Zhang", "Yunta Hsieh", "Qi Han", "Zhongwei Wan", "Ziheng Zhang", "Jingxuan Zhang", "Jing Xiong", "Ziyuan Liu", "Yifan Zhang", "Hangrui Cao", "Chenyang Zhao", "Mi Zhang" ]	[ "cs.CV" ]	[ "Computer Science" ]	2026-03-16T00:00:00	https://arxiv.org/abs/2603.14989	https://arxiv.org/pdf/2603.14989v1	2603.14989	10.48550/arXiv.2603.14989	1	false	null	arXiv.org	0.5099

End of preview. Expand in Data Studio

Speculative Decoding Papers — FineSet

A research-paper dataset on Speculative Decoding Papers, assembled, deduplicated, and quality-scored by FineSet from arXiv and Semantic Scholar.

📸 This is a dated snapshot — generated 2026-06-19. It is not auto-updated. Research on Speculative Decoding Papers moves fast — new papers land on arXiv every week. Want this same dataset refreshed daily, on a topic you choose? See the bottom. ↓

Why this dataset

Quality-scored: quality_score float (0–1), blends citations with recency + code/venue signals — filter out the noise
Papers with code: 134 flagged via has_code — find reproducible work fast
Deduplicated: arXiv + Semantic Scholar cross-referenced, duplicate records merged
Clean JSONL: 485 records, one per line, normalized fields — no encoding garbage

Dataset details

Records: 485
Date range: 2022–2026
Snapshot date: 2026-06-19 (frozen — see note above)
Sources: arXiv, Semantic Scholar (cross-referenced, duplicates merged)
arXiv categories: cs.LG, cs.CL
Quality scoring: citations + recency + code/venue blend, 0–1 (p50=0.35, p90=0.61)
Format: JSONL, one record per line

Fields

Field	Type	Description
id	string	Deterministic SHA256 record id
sources	list	Which sources contributed (`arxiv`, `semantic_scholar`)
title	string	Paper title
abstract	string	Full abstract
authors	list	Author names
categories	list	arXiv category codes
fields_of_study	list	Semantic Scholar field tags
published_date	string	ISO 8601 date
url	string	arXiv abstract URL
pdf_url	string\|null	Open-access PDF if available
arxiv_id	string\|null	arXiv identifier
doi	string\|null	DOI if available
citation_count	int	Citation count (Semantic Scholar)
influential_citation_count	int	Influential citations (Semantic Scholar)
has_code	bool	Code repo detected in the arXiv comment
code_url	string\|null	GitHub URL if detected
venue	string\|null	Publication venue
quality_score	float	0–1, blended (citations + recency + code/venue)

Quality score methodology

quality_score = max(impact, freshness), clamped to [0, 1], where:

impact = max( log10(citations+1)/4 , log10(influential_citations+1)/2 ) — realized impact (0.5 at 100 citations, ~0.75 at 1,000, 1.0 at 10,000+).
freshness = recency × (0.35 + 0.30·has_code + 0.20·has_venue) — a baseline for recent papers (so a strong paper published this week isn't scored 0 just for lacking citations), where recency is 1.0 for papers ≤60 days old and decays linearly to 0 by ~18 months.

Old highly-cited papers score on impact; brand-new papers score on freshness; old uncited papers score ~0. Useful for filtering training data by quality, not just age.

👉 Want this on YOUR topic, updated daily?

This snapshot is frozen at 2026-06-19. The live FineSet pipeline keeps a dataset like this refreshed every day on whatever topic you describe — new papers in, dedup and quality scoring automatic, export as JSONL/Parquet or push straight to the Hub.

Tell me the topic you'd want and I'll run the pipeline on it — open a discussion on this dataset, it's free and it's how I decide what to build next.

→ fineset.io — describe what you want to train on, get a dataset. Early-access waitlist open (referral skip available).

Downloads last month: 31