[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84224-en":3,"doc-seo-84224-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},84224,962075114101,"Seraphina","https://ap-avatar.wpscdn.com/avatar/e000253a75eb197efd?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780044092746381165",8,"Research & Report","DeLS-Spec Decoupled Long-Short Contexts for Parallel Speculative Drafting","Speculative decoding speeds up LLM inference by drafting multiple tokens and verifying them in parallel with the target model. Block-parallel drafters like DFlash predict blocks efficiently but largely miss explicit intra-block causal conditioning, which can reduce acceptance. DeLS-Spec decouples long-context and short-context modeling by keeping DFlash fixed as a long-context expert and adding a lightweight local head for short-context causality. The local head trains independently, with low cost, and is combined with DFlash logits during inference. Experiments on Qwen3 show consistent gains in speedup and average acceptance length across math, code, and dialogue benchmarks.","arXiv :2607 .07409v 1 [ cs .CL] 8 Jul 2026  \nDELS-SPEC: DECOUPLED LONG-SHORT CONTEXTS FOR PARALLEL SPECULATIVE DRAFTING  \nHong-Kai Zheng, Piji Li ∗  \nCollege of Artificial Intelligence,  \nNanjing University of Aeronautics and Astronautics, China  \nMIIT Key Laboratory of Pattern Analysis and Machine Intelligence, Nanjing, China The Key Laboratory of Brain-Machine Intelligence Technology,  \nMinistry of Education, Nanjing, China {dt ttt,[pjli](pjli}@nuaa.edu.cn)[}](pjli}@nuaa.edu.cn)[@nuaa.edu.cn](pjli}@nuaa.edu.cn)  \nABSTRACT  \nSpeculative decoding accelerates LLM inference by drafting multiple tokens and verifying them in parallel. Block-parallel drafters such as DFlash further improve drafting efficiency by predicting an entire block in one pass, but their position-wise predictions lack explicit intra-block causal conditioning. Recent methods such as Domino and DSpark attempt to introduce such causality into block-parallel drafting, but they require training the draft model from scratch, which limits their flexibility and increases training cost. We propose DeLS-Spec, a decoupled long-short context speculative decoding method. DeLS-Spec treats the fixed DFlash model as a long-context expert and introduces a lightweight local head as a short-context expert. The local head can be trained independently with a standard next-token prediction objective, without joint training with the target model or the DFlash backbone, leading to extremely low training cost. At inference time, DeLSSpec combines long-context and short-context logits, and the local head is not tied to a specific DFlash checkpoint, making the method more modular and flexible. Experiments on Qwen3 models show that DeLS-Spec consistently improves speedup and average acceptance length over DFlash across math, code, and dialogue benchmarks. Code is available at GitHub.  \n1 INTRODUCTION  \nLarge language models are typically decoded autoregressively (Achiam et al., 2023; Liu et al., 2024; Yang et al., 2025), where each generated token requires a new forward pass of the target model. This sequential process becomes a major latency bottleneck, especially for long-form generation and interactive applications. Speculative decoding mitigates this problem by using a lightweight draft model to propose multiple future tokens, which are then verified in parallel by the target model while preserving the target distribution (Leviathan et al., 2023; Chen et al., 2023) .  \nThe efficiency of speculative decoding depends heavily on the drafter (Cai et al., 2024; Ankner et al., 2024; Li et al., 2024a;b; 2026b) . Traditional autoregressive drafters maintain strong local consistency but still generate draft tokens sequentially (Li et al., 2024a; Cheng et al., 2024) . Recent block-parallel drafters, such as DFlash, improve drafting efficiency by predicting an entire block of tokens in one pass (An et al., 2025; Liu et al., 2026; Chen et al., 2026) . However, this parallelism also introduces a limitation: tokens inside the same draft block are predicted largely independently, without explicitly conditioning on previously drafted tokens in that block (Chen et al., 2026; Huang et al., 2026; Rheinboldt et al., 2026) . As a result, DFlash provides strong long-context predictions from the prefix, but lacks explicit short-context causal modeling within the draft block.  \nSeveral recent methods, including Domino and DSpark, attempt to address this issue by introducing intra-block causality into block-parallel drafting (Huang et al., 2026; Cheng et al., 2026b; Hu et al., 2026; Rheinboldt et al., 2026) . These methods show that modeling causal dependencies inside the  \n∗ Corresponding author  \ndraft block can improve acceptance. However, they usually require training a new draft model from scratch or jointly training the draft backbone with additional causal components. This makes them costly to apply when a DFlash-style drafter has already been trained, and limits their flexibility across different d","cbCaio1EE8BFaZYN","https://ap.wps.com/l/cbCaio1EE8BFaZYN","pdf",517146,1,17,"English","en",105,"# Introduction\n## Autoregressive decoding and speculative decoding\n## Block-parallel drafting limitations\n## DeLS-Spec approach and decoupled training\n## Experimental evaluation and contributions","[{\"question\":\"What problem does DeLS-Spec address in block-parallel speculative decoding?\",\"answer\":\"Block-parallel drafters like DFlash predict tokens within a draft block with limited explicit intra-block causal conditioning, so the method improves long-context prediction from the prefix but weakens short-context causal modeling inside the block.\"},{\"question\":\"How does DeLS-Spec combine long-context and short-context information during inference?\",\"answer\":\"DeLS-Spec treats the fixed DFlash model as a long-context expert and uses a lightweight local head as a short-context expert, then combines their long-context logits and short-context logits to guide acceptance and generation.\"},{\"question\":\"Why is training the local head in DeLS-Spec considered low cost?\",\"answer\":\"The local head is trained independently using a standard next-token prediction objective on plain text, without requiring target-model hidden states, DFlash hidden states, or joint optimization with the speculative decoding pipeline.\"}]",1784194151,43,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"dels-spec-decoupled-long-short-contexts-for-parallel-speculative-drafting","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/dels-spec-decoupled-long-short-contexts-for-parallel-speculative-drafting/84224/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does DeLS-Spec address in block-parallel speculative decoding?","Question",{"text":75,"@type":76},"Block-parallel drafters like DFlash predict tokens within a draft block with limited explicit intra-block causal conditioning, so the method improves long-context prediction from the prefix but weakens short-context causal modeling inside the block.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does DeLS-Spec combine long-context and short-context information during inference?",{"text":80,"@type":76},"DeLS-Spec treats the fixed DFlash model as a long-context expert and uses a lightweight local head as a short-context expert, then combines their long-context logits and short-context logits to guide acceptance and generation.",{"name":82,"@type":73,"acceptedAnswer":83},"Why is training the local head in DeLS-Spec considered low cost?",{"text":84,"@type":76},"The local head is trained independently using a standard next-token prediction objective on plain text, without requiring target-model hidden states, DFlash hidden states, or joint optimization with the speculative decoding pipeline.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]