[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"topic-zh-google-turboquant-llm-":3},{"cluster":4,"timeline":17},{"id":5,"slug":6,"title":7,"pinned":8,"status":9,"summary":10,"category":11,"language":12,"created_at":13,"merged_into":14,"article_count":15,"first_seen_at":13,"last_updated_at":16},"aad23a48-1a6a-4238-b583-f5673eda0fcb","google-turboquant-llm-","Google TurboQuant 壓低 LLM 記憶體成本",false,"active","Google 推出 TurboQuant，結合 QJL 與 PolarQuant，主打壓低 vector quantization 的記憶體開銷，並宣稱 LLM inference 最高可快 8 倍。","research","zh","2026-07-05T09:01:57.363435+00:00",null,4,"2026-07-06T00:40:00.034514+00:00",[18,25,32],{"id":19,"slug":20,"title":21,"summary":22,"category":11,"image_url":23,"cover_image":23,"published_at":24,"is_canonical_seed":8},"1ac8d1f1-a68e-41d3-9955-d3c3a6b55e39","turboquant-online-vector-quantization-near-optimal-zh","TurboQuant：線上向量量化更接近最優","TurboQuant 提出可線上運作、適合加速器的向量量化方法，主打 MSE 與 inner-product 失真都能逼近近似最優，並鎖定 KV cache 與向量搜尋場景。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777467660452-rlcy.png","2026-04-29T13:00:39.298+00:00",{"id":26,"slug":27,"title":28,"summary":29,"category":11,"image_url":30,"cover_image":30,"published_at":31,"is_canonical_seed":8},"82766fdc-4368-445d-bb4a-03377726df02","turboquant-cuts-memory-use-without-accuracy-loss-zh","TurboQuant 省 6 倍記憶體，還不掉準確率","Google Research 發表 TurboQuant，主打記憶體用量降到 1\u002F6、推論快 8 倍，且在報告測試中沒有準確率損失。這篇看它怎麼改 AI 伺服器成本。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775161134112-ftrj.png","2026-04-02T20:18:39.073+00:00",{"id":33,"slug":34,"title":7,"summary":10,"category":11,"image_url":35,"cover_image":35,"published_at":36,"is_canonical_seed":37},"6ea121bb-a78e-4bc2-bda3-9be1e048ab95","googles-turboquant-cuts-llm-memory-costs-zh","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775160769707-5e2g.png","2026-04-02T20:12:31.746+00:00",true]