[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-7518634b-1961-4484-8047-9f787dcb2c18":3},{"id":4,"title":5,"summary":6,"original_url":7,"source_id":8,"tags":9,"published_at":20,"created_at":21,"modified_at":22,"is_published":23,"publish_type":24,"image_url":13,"view_count":25},"7518634b-1961-4484-8047-9f787dcb2c18","DeLS-Spec 把 DFlash 拆成「长-短双专家」:加一个本地头就能再提速","DFlash 把整块一次性起草,把投机解码效率抬了一截,但块内每个位置缺显式因果;Domino、DSpark 想补短板,代价却是从头重训草稿模型。\n\n7 月 8 日挂在 arXiv 的 DeLS-Spec(arXiv:2607.07409) 给出更轻的路子:把现成 DFlash 冻住当「长上下文专家」,再额外训练一个轻量本地头作为「短上下文专家」。本地头只用标准 next-token 目标独立训练,既不要联合训练目标模型,也不绑定特定 DFlash 版本,训练成本几乎可忽略。推理时两路 logits 融合:长程依赖由 DFlash 兜底,块内一致性由本地头补齐。在 Qwen3 上的实验里,DeLS-Spec 在数学、代码、对话三类基准上一致超过 DFlash 原版,平均接受长度同步抬升。\n\n比起 Domino\u002FDSpark「重训一切」的思路,DeLS-Spec 更像插件式改造——对已经部署 DFlash 的团队几乎零迁移成本。它的价值不在炫技式的端到端重训,而在于给出高度模块化的设计模板:长程能力交给已成熟的专家,短程一致性用极轻的本地头补齐。这条「冻结主模型 + 插件子头」的范式,或许会被接下来一连串推理加速工作借用。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.07409","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17],{"id":11,"name":12,"slug":12,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"id":18,"name":19,"slug":19,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm","2026-07-10T00:08:00Z","2026-07-10T00:11:07.410278Z","2026-07-10T00:11:07.410293Z",true,"agent",3]