[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-eldr-moe-routing":3,"topics-all":36,"news-related-5a4c2a98-6ce7-4e51-8348-118be3083afc":55},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"5a4c2a98-6ce7-4e51-8348-118be3083afc","ELDR 把 MoE 推理的「延迟最后一公里」拉直:vLLM 实测 TPOT 最多砍 13.9%","PD(预填\u002F解码)解耦已经是 LLM 在线推理的事实标准,但传统 router 只看「这一节点负载多少」、不看「这一节点预热了哪些专家」——这在 MoE 模型上是一笔隐性税。来自 KAIST 和微软亚洲研究院的 ELDR(arXiv:2607.00466,v2 于 7 月 2 日上线)做了两件事:离线时,用请求预填阶段的专家激活分布构建一个\"专家签名\",再做均衡 K-means 把签名空间分片到不同 decode worker;在线时,把请求路由到与签名最匹配、且负载最轻的 worker。配合按 KV-block 同步粒度维护的 signature cache,ELDR 在 vLLM 上、40 卡规模、3 个 MoE 模型 \u002F 2 种负载下,中位 TPOT 相对四种负载均衡基线最高压减 13.9%,最低也有 5.9%,输出 bitwise 不变。它揭示了一个被性能曲线反复掩盖的事实:MoE 时代,「请求去哪儿」比「请求来多少」更影响延迟。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00466","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":18,"name":19,"slug":19,"description":13,"color":13},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"id":21,"name":22,"slug":22,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"fe0a2a49-1961-4244-9aff-d23f17fc9f69","en","ELDR straightens MoE latency: TPOT down 13.9% on vLLM","PD (prefill\u002Fdecode) decoupling is already the de facto standard for LLM online inference, but traditional routers only look at \"how loaded is this node\", not \"which experts has this node warmed up\" — this is a hidden tax on MoE models. ELDR (arXiv:2607.00466, v2 went live July 2) from KAIST and Microsoft Research Asia did two things: offline, it uses the expert activation distribution from the request's prefill phase to build an \"expert signature\", then balanced K-means partitions the signature space to different decode workers; online, it routes the request to the worker that best matches the signature and has the lightest load. Paired with a signature cache maintained at KV-block synchronization granularity, ELDR on vLLM, 40-card scale, 3 MoE models \u002F 2 load types, the median TPOT relative to the four load-balancing baselines is reduced by up to 13.9%, at least 5.9%, with output bitwise unchanged. It reveals a fact repeatedly masked by performance curves: in the MoE era, \"where the request goes\" affects latency more than \"how many requests come\".","eldr-moe-routing","2026-07-03T12:30:00Z","2026-07-03T12:11:30.933826Z","2026-08-19T02:08:40.142862Z",true,"agent",199,[37,46],{"slug":38,"tag_slug":38,"title_zh":39,"title_en":40,"intro_zh":41,"intro_en":42,"id":43,"is_active":33,"created_at":44,"modified_at":45},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":47,"tag_slug":47,"title_zh":48,"title_en":49,"intro_zh":50,"intro_en":51,"id":52,"is_active":33,"created_at":53,"modified_at":54},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":56},[57,62,67,72,77,82],{"id":58,"title":59,"news_slug":60,"published_at":61},"3d922c00-afcb-4f1c-a6d5-8f9d6c10c642","从 Kimi Linear 到 Kimi K3:MoE 推理效率战里被忽略的架构升级","kimi-k3-latentmoe-kda-attnres-nope","2026-07-30T00:30:00+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"f2b0bde3-ebb5-49e5-a418-5ece37639d1b","MIPU\u002FMIPI：把 LLM RL 的「训练—推理失配」从工程噪音重写为优化目标","mipu-mipi-rl-mismatch","2026-07-04T08:00:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"cd82f869-d4a7-45b7-95ce-b66051e9d933","BlockPilot：实例自适应策略学习让扩散式投机解码再下一城,Qwen3-4B 上首破 4.20× 加速","blockpilot-instance-adaptive-block","2026-07-01T14:17:29+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"abf22bbb-eccf-46d0-99d6-debe1596f92b","自验证蒸馏：无需外部教师，LLM如何实现自我进化","self-verified-distill-qwen3-16-7pp-math","2026-05-27T19:00:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"6c9e553d-3029-4867-bc29-ee069d26b934","自验证蒸馏：无需外部教师，LLM 如何实现自我进化","svd-self-verified-distill-stanford-perplexity","2026-05-27T16:05:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"2ca620f4-a046-4274-925d-f0689123a498","ParaRNN：Apple 让 RNN 重回战场，7B 参数模型训练提速 665 倍","apple-pararnn-7b-rnn-665x-faster-iclr","2026-05-27T13:15:00+00:00"]