[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-rag17-sod1-als-nature-medicine-template-zh":3,"article-related-rag17-sod1-als-nature-medicine-template-zh":31,"series-research-584d0e47-4c7e-469e-8d28-236966c00188":74},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":23,"views":27,"created_at":28,"published_at":29,"topic_cluster_id":30},"584d0e47-4c7e-469e-8d28-236966c00188","rag17-sod1-als-nature-medicine-template-zh","RAG-17 把 SOD1-ALS 寫成可抄模板","\u003Cp data-speakable=\"summary\">以前是把動物數據和早期人體結果分開講，現在是把它們排成一條能直接抄的轉譯故事線。\u003C\u002Fp>\u003Cp>我最近一直在看各種 biotech 發稿，老實說，大多數都像是三個部門各寫一段，最後硬拼成一篇。前臨床講得很滿，人體數據講得很保守，兩邊都像在怕被 reviewer 盯。我最煩的就是這種：看起來很熱鬧，實際上沒把「這個東西到底怎麼從老鼠走到人」講清楚。\u003C\u002Fp>\u003Cp>Ractigen 的 \u003Ca href=\"\u002Ftag\u002Frag\">RAG\u003C\u002Fa>-17 這篇我會多看一眼，因為它至少把 delivery、動物\u003Ca href=\"\u002Fnews\u002Fsurvey-of-large-language-models-zh\">模型\u003C\u002Fa>、Phase 1 biomarker 放在同一條線上。觸發我去拆它的來源是這則 \u003Ca href=\"https:\u002F\u002Fwww.biospace.com\u002Fpress-releases\u002Fractigen-therapeutics-announces-landmark-publication-in-nature-medicine-highlighting-unprecedented-preclinical-efficacy-and-positive-first-in-human-clinical-data-for-rag-17-in-sod1-als\" target=\"_blank\" rel=\"noopener noreferrer\">BioSpace\u003C\u002Fa> 轉載，原始稿在 \u003Ca href=\"https:\u002F\u002Fwww.prnewswire.com\u002Fnews-releases\u002Fractigen-therapeutics-announces-landmark-publication-in-nature-medicine-highlighting-unprecedented-preclinical-efficacy-and-positive-first-in-human-clinical-data-for-rag-17-in-sod1-als-302826429.html\" target=\"_blank\" rel=\"noopener noreferrer\">PR Newswire\u003C\u002Fa>，也指向 \u003Ca href=\"https:\u002F\u002Fwww.nature.com\u002Fnaturemedicine\u002F\" target=\"_blank\" rel=\"noopener noreferrer\">Nature Medicine\u003C\u002Fa> 的論文頁面。來源沒有提供觀看數或書籤數，我就不亂補。\u003C\u002Fp>\u003Ch2>我先看 delivery，因為 CNS RNA 藥最常死在這裡\u003C\u002Fh2>\u003Cblockquote>“RAG-17 utilizes Ractigen's proprietary Smart Chemistry-Aided Delivery (SCAD™) technology, which conjugates the siRNA duplex to a specialized accessory oligonucleotide (ACO). This allows for broad distribution throughout the central nervous system (CNS) and highly durable gene silencing following intrathecal (IT) injection.”\u003C\u002Fblockquote>\u003Cp>翻譯一下就是：Ractigen 不只是在賣 siRNA，它是在賣一套把 siRNA 送進 CNS 的方法。這很重要，因為 CNS RNA therapeutics 的麻煩從來不是「有沒有靶點」，而是「送不送得到、留不留得住、會不會每次打藥都像在做侵入式手術」。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784678591746-wzd8.png\" alt=\"RAG-17 把 SOD1-ALS 寫成可抄模板\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>我看過太多 program，靶點漂亮得要命，結果 delivery 爛到不行。payload 進不去，進去了也不夠久，最後整個故事都在講 chemistry 的想像力，不是在講產品。這篇的重點其實很明白：SCAD™ 是中心，不是配角。它聲稱 ACO \u003Ca href=\"\u002Fnews\u002Fpersona-steering-llm-capabilities-analysis-zh\">會改變\u003C\u002Fa>分布和持續時間，而這正是我會先驗證的地方。\u003C\u002Fp>\u003Cp>實操寫法很簡單。你在看任何早期療法稿件時，先把疾病名拿掉，只問三件事：能不能到 tissue、能不能維持足夠久、給藥方式能不能被臨床接受。只要這三題有一題講不清楚，後面的漂亮字眼通常都只是裝飾。\u003C\u002Fp>\u003Cul>\u003Cli>先判斷 delivery bottleneck，再看 target biology。\u003C\u002Fli>\u003Cli>把 route-of-administration 和機制分開讀。\u003C\u002Fli>\u003Cli>優先找 durability，不要只看 peak knockdown。\u003C\u002Fli>\u003C\u002Ful>\u003Cp>如果是 CNS RNAi，delivery 不是註腳，它就是產品本身。這也是為什麼我會把平台放在疾病之前寫，因為讀者真正要知道的是：這套 chemistry 到底有沒有把「送藥」這件事做成可重複的方法。\u003C\u002Fp>\u003Ch2>生物標記比較誠實，勝利宣言比較吵\u003C\u002Fh2>\u003Cblockquote>“In Cohort 1, mean cerebrospinal fluid (CSF) SOD1 protein decreased by 69% at Day 240, while plasma NfL… decreased by a mean of 62%, with individual nadirs reaching up to 85% below baseline.”\u003C\u002Fblockquote>\u003Cp>這段的意思是，它想證明的不只是安全，而是藥理真的有在動。這個方向是對的。對 SOD1-ALS 這種病來說，CSF SOD1 是直接機制讀出，NfL 則比較像神經傷害的總體指標。兩個一起看，比只丟一個 target protein 好很多。\u003C\u002Fp>\u003Cp>我以前看過一個 CNS program，團隊只拿 target engagement 來講，完全沒補 injury marker。結果會議上每個人都在猜：你是把分子打下去了，還是真的讓病程有變化？那種簡報最容易讓人失焦。RAG-17 這裡比較聰明，因為 CSF SOD1 和 plasma NfL 至少把「機制有沒有碰到」和「神經是不是還在受損」分開處理。\u003C\u002Fp>\u003Cp>實操上，我會建議你在寫 early clinical update 時，永遠至少放兩層 biomarker。第\u003Ca href=\"\u002Fnews\u002Fagent-skills-llm-agents-next-layer-zh\">一層\u003C\u002Fa>是 target 層，第二層是 disease burden 或 tissue injury 層。只有第一層，故事很容易變成「分子變了所以應該有用」；有第二層，才有辦法把 translational claim 寫得像樣一點。\u003C\u002Fp>\u003Cul>\u003Cli>一個 biomarker 看機制，一個看疾病影響。\u003C\u002Fli>\u003Cli>把時間點寫清楚，不要只寫幅度。\u003C\u002Fli>\u003Cli>平均值和個體最低點要分開講。\u003C\u002Fli>\u003C\u002Ful>\u003Cp>我也要提醒一句，69% 這種數字很會騙人。你得問：什麼時候降的、能維持多久、baseline 波動有多大。沒有這些，數字只是數字。但作為轉譯包裝，這組 biomarker 至少是有骨架的，不是空話。\u003C\u002Fp>\u003Ch2>安全性很無聊，直到它決定這案子能不能活下去\u003C\u002Fh2>\u003Cblockquote>“RAG-17 met its primary safety endpoint. It was well-tolerated, with no serious adverse events (SAEs) and no requirement for invasive mechanical ventilation up to the data cutoff.”\u003C\u002Fblockquote>\u003Cp>這句話的白話版是：第一輪人體資料沒有炸掉。這聽起來很平淡，但在 ALS 這種病裡，平淡常常就是好消息。病人和醫師可以接受風險，但不能接受你一邊想治病，一邊又製造另一個更麻煩的問題。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784678590541-o6vu.png\" alt=\"RAG-17 把 SOD1-ALS 寫成可抄模板\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>很多 biotech 寫法會把 safety 當成一段沒人想看的背景資料。我完全不這樣看。安全性是早期 program 的門票，沒有這關，後面的 biomarker 再漂亮也只是紙上談兵。這裡至少有幾個我在意的點：沒有 SAEs、沒有侵入式機械通氣需求，這代表 program 沒有在第一個關卡把自己搞死。\u003C\u002Fp>\u003Cp>但我也不會過度浪漫化。樣本小、時間短、追蹤還早，這些限制都在。它現在比較像是「沒看到明顯紅旗」，不是「可以放心了」。如果我是內部簡報，我會這樣寫：安全性過關，但還不到可以放鬆的程度。\u003C\u002Fp>\u003Cp>實操寫法是：先講 primary safety endpoint，再用具體語句定義 tolerated 到底是什麼。沒有 SAEs 很好；沒有侵入式通氣更好；但 sample size 也要一起亮出來。六個人是 signal，不是結論。\u003C\u002Fp>\u003Cp>這也是外部讀者最愛過度解讀的地方。看到 no SAEs 就直接腦補成快要核准，這種跳法很危險。它只是說這個案子先活過第一道門，還沒證明後面一定會成。\u003C\u002Fp>\u003Ch2>晚期動物救援有說服力，但你得先看時間點\u003C\u002Fh2>\u003Cblockquote>“In aggressively progressing SOD1 G93A mouse models, RAG-17 demonstrated remarkable efficacy even when administered significantly after symptom onset… Late-stage treatment extended survival by up to 75.8%.”\u003C\u002Fblockquote>\u003Cp>這段真正有價值的地方，不是 “remarkable” 這種形容詞，而是 dosing timing。它不是只在很早期、很乾淨的模型裡做預防，而是在症狀已經開始後還能看到效果。對轉譯工作來說，這比單純的預防性實驗更難糊弄人。\u003C\u002Fp>\u003Cp>我對「unprecedented」這種字眼一向很敏感，因為太多動物數據在進入真實世界後就掉下去。真正值得看的不是形容詞，是模型設計。晚期給藥、快速惡化的 ALS 模型，這些條件都讓結果比較難被輕鬆打發。\u003C\u002Fp>\u003Cp>同樣的邏輯也適用在 rat 和 NHP。rat 看的是疾病延遲和脊髓 motor neuron 保留；cynomolgus monkey 則看 lumbar spinal cord 的 knockdown 是否真的持久。這些東西湊在一起，才比較像跨物種的藥理一致性，不是單一模型的偶然勝利。\u003C\u002Fp>\u003Cp>實操上，我會建議你每次看到漂亮的 animal efficacy，都固定問三件事：是預防還是治療？離 onset 有多晚？有沒有第二物種或更像人的 tissue context？如果答案都沒有，別急著把 mouse win 寫成 human promise。\u003C\u002Fp>\u003Cul>\u003Cli>晚期給藥比早期預防更有轉譯價值。\u003C\u002Fli>\u003Cli>跨物種一致性比單一模型更有說服力。\u003C\u002Fli>\u003Cli>持續時間比單次驚人終點更重要。\u003C\u002Fli>\u003C\u002Ful>\u003Cp>我不是說動物資料就能證明人體有效。當然不能。但 timing 和 durability 至少讓這份 package 比一般那種「我們治好了老鼠」的稿子可信很多。\u003C\u002Fp>\u003Ch2>NHP bridge 是很多 RNA program 最愛跳過的那一段\u003C\u002Fh2>\u003Cblockquote>“In cynomolgus monkeys, intrathecal RAG-17 achieved up to 91% reduction of SOD1 mRNA in the lumbar spinal cord, an effect that persisted for up to 72 days post-dose.”\u003C\u002Fblockquote>\u003Cp>這句的意思很直接：它不是只在鼠類裡面有效，還有一段接近人體的橋接資料。對 RNA 藥來說，NHP target engagement 很重要，因為它逼你面對分布、持續性、組織相關性這三件事，沒辦法再靠漂亮敘事混過去。\u003C\u002Fp>\u003Cp>我會特別看 primate data，就是因為它很難包裝成「大概差不多」。如果藥到不了對的位置、knockdown 不夠久，或給藥方式太麻煩，故事就會開始卡。這裡 72 天的持續性，至少支持了較低給藥頻率的說法，這對病人和臨床都不是小事。\u003C\u002Fp>\u003Cp>如果你在做轉譯敘事，我建議一定要放一個接近人體 anatomy 或 physiology 的 bridge species。然後把 duration 寫成天數，不要寫成「持久」這種空話。天數是誠實的，形容詞通常不是。\u003C\u002Fp>\u003Cp>實操寫法就是：在 rodent efficacy 和 human data 中間，補一段能證明 delivery 與 target engagement 的 NHP 資料。再把 dose frequency、exposure-response、以及 lumbar spinal cord 跟臨床相關 compartment 的對應關係交代清楚。講得出來就寫，講不出來就承認還缺一塊。\u003C\u002Fp>\u003Cp>這段是很多 program 最常省略的，但我反而覺得它最值錢。因為它把「動物有效」和「人類可行」之間那段最難看的距離，硬生生補了一部分。\u003C\u002Fp>\u003Ch2>我喜歡這份稿子的地方，是它有順序\u003C\u002Fh2>\u003Cp>Ractigen 這次最像樣的地方，不是用了多少大字，而是它的順序對了。它先講 platform chemistry，再講動物 rescue，再講 primate target engagement，最後才進到 first-in-human biomarker 和 safety。這條路線才像一份正常的 translational story，不像三篇不同稿子硬貼在一起。\u003C\u002Fp>\u003Cp>我對那種一上來就從 mouse survival 跳到 “best-in-class” 的寫法真的很沒耐心。這篇至少有試著把路補完整：delivery 做什麼、動物看到了什麼、NHP 怎麼橋接、人體先看到什麼、還缺什麼。這樣寫，讀者才知道你是在描述證據，不是在賣夢。\u003C\u002Fp>\u003Cp>實操上，你寫任何早期 CNS RNA program，都可以照這個順序來：先平台，再最佳動物模型，再 bridge species，再人體早期資料，最後補限制。順序一亂，大家就會懷疑你是不是在藏東西。\u003C\u002Fp>\u003Cp>我也想提醒一個很常見的壞習慣：把 signal 和 proof 混在一起。這篇的 signal 是有的，甚至不算弱；但它還不是 proof。這句話一定要留著，因為早期資料最怕的就是被寫成已經證明一切。\u003C\u002Fp>\u003Cp>如果你是要寫給開發者或研究者看，最有價值的不是把話說滿，而是把證據層級標清楚。這才是能拿去討論、拿去決策的版本。\u003C\u002Fp>\u003Ch2>可抄的模板\u003C\u002Fh2>\u003Cpre>\u003Ccode># Translational announcement template for an RNA therapeutic in CNS disease\n\n## Headline\n[Company] announces [journal publication \u002F clinical milestone] for [asset] in [disease]\n\n## One-line summary\n[Asset] shows [mechanism] in [target tissue], with [preclinical outcome] and [early clinical biomarker\u002Fsafety outcome].\n\n## Platform paragraph\n[Asset] uses [delivery platform \u002F chemistry] to reach [tissue], sustain [target engagement], and reduce [dosing burden].\n\n## Preclinical package\n- Model 1: [species\u002Fmodel], [dose], [timing relative to disease onset], [primary outcome]\n- Model 2: [species\u002Fmodel], [secondary outcome], [durability]\n- Bridging species: [NHP\u002Fspecies], [target engagement metric], [duration of effect]\n\n## Human data\n- Study design: [open-label \u002F randomized \u002F dose-escalation]\n- Population: [n], [disease subtype], [inclusion criteria]\n- Safety: [SAEs yes\u002Fno], [common TEAEs], [tolerability summary]\n- Pharmacodynamics: [target biomarker], [injury biomarker], [timepoint], [percent change]\n- Exploratory clinical readouts: [function \u002F respiratory \u002F other], [direction of change]\n\n## Interpretation paragraph\nThese data suggest [mechanistic conclusion] and support further study of [asset] as a [disease-modifying \u002F symptom-modifying] therapy. The strongest claim is [what the data actually support]. The weakest claim is [what the data do not yet prove].\n\n## Copy-ready caution language\nThese findings are based on [small sample size \u002F preclinical models \u002F early follow-up] and should be interpreted as [signal \u002F hypothesis-generating evidence], not confirmation of clinical benefit.\n\n## Practical takeaway\nIf you are writing about an early CNS RNA program, lead with delivery, then durability, then biomarker movement, then safety, then the limits.\n\u003C\u002Fcode>\u003C\u002Fpre>\u003Cp>這段可以直接拿去改。你只要把方括號補滿，就能把一份前臨床加 Phase 1 的混合資料，整理成一篇不會太浮誇、但也不會散掉的 translational write-up。\u003C\u002Fp>\u003Cp>我會特別建議 RNAi、ASO、或任何 CNS oligo program 都用這個骨架。因為這類案子最容易被寫成「化學很厲害、結果很模糊」，而這個模板至少能逼你把 delivery、durability、biomarker、safety 一層一層拆開。\u003C\u002Fp>\u003Cp>來源致謝：我這篇是根據 \u003Ca href=\"https:\u002F\u002Fwww.biospace.com\u002Fpress-releases\u002Fractigen-therapeutics-announces-landmark-publication-in-nature-medicine-highlighting-unprecedented-preclinical-efficacy-and-positive-first-in-human-clinical-data-for-rag-17-in-sod1-als\" target=\"_blank\" rel=\"noopener noreferrer\">BioSpace\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fwww.prnewswire.com\u002Fnews-releases\u002Fractigen-therapeutics-announces-landmark-publication-in-nature-medicine-highlighting-unprecedented-preclinical-efficacy-and-positive-first-in-human-clinical-data-for-rag-17-in-sod1-als-302826429.html\" target=\"_blank\" rel=\"noopener noreferrer\">PR Newswire 原文\u003C\u002Fa>，以及其指向的 \u003Ca href=\"https:\u002F\u002Fwww.nature.com\u002Fnaturemedicine\u002F\" target=\"_blank\" rel=\"noopener noreferrer\">Nature Medicine\u003C\u002Fa> 資訊整理而成；我的拆解方法、段落順序和模板是原創。","我拆 Ractigen 的 RAG-17 故事，順手整理成一個能直接套用的前臨床＋Phase 1 轉譯模板。","www.biospace.com","https:\u002F\u002Fwww.biospace.com\u002Fpress-releases\u002Fractigen-therapeutics-announces-landmark-publication-in-nature-medicine-highlighting-unprecedented-preclinical-efficacy-and-positive-first-in-human-clinical-data-for-rag-17-in-sod1-als",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784678591746-wzd8.png","research","zh","84f909d5-e578-49ad-9f8e-c8cafc7562ea",[17,18,19,20,21,22],"RAG-17","SOD1-ALS","siRNA","CNS delivery","biomarker","Phase 1",[24,25,26],"先看 delivery，再看 disease，CNS RNA 藥的重點是能不能送到、留得住。","雙 biomarker 比單一 target readout 更能支撐轉譯敘事。","可抄的模板要把 platform、animal、NHP、human data、limitations 按順序寫清楚。",0,"2026-07-22T00:02:47.756002+00:00","2026-07-22T00:02:47.743+00:00","3f294285-b3a6-4956-bd42-12d23fcba798",{"tags":32,"relatedLang":33,"relatedPosts":37},[],{"id":15,"slug":34,"title":35,"language":36},"rag17-sod1-als-nature-medicine-template-en","RAG-17 turns SOD1-ALS data into a template","en",[38,44,50,56,62,68],{"id":39,"slug":40,"title":41,"cover_image":42,"image_url":42,"created_at":43,"category":13},"f039531b-dbe8-43e5-a037-5ad6ca590524","survey-of-large-language-models-zh","大型語言模型全景整理","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784629980760-ftkd.png","2026-07-21T10:32:29.369537+00:00",{"id":45,"slug":46,"title":47,"cover_image":48,"image_url":48,"created_at":49,"category":13},"55d40b40-0d7a-4ffb-906b-18b284fb3a3a","evaluating-memory-in-llm-agents-zh","用多輪互動測 LLM 記憶","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784628186159-2km8.png","2026-07-21T10:02:36.154394+00:00",{"id":51,"slug":52,"title":53,"cover_image":54,"image_url":54,"created_at":55,"category":13},"828339d3-50c4-47fd-ba13-1a50f8430793","persona-steering-llm-capabilities-analysis-zh","Persona steering 會改變模型能力嗎","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784626389337-g5ex.png","2026-07-21T09:32:27.763156+00:00",{"id":57,"slug":58,"title":59,"cover_image":60,"image_url":60,"created_at":61,"category":13},"331ebfe2-bbcb-4e5f-be0a-043310c0a710","llm-inference-hardware-memory-interconnect-zh","LLM 推理瓶頸不在算力","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784622786324-dwuy.png","2026-07-21T08:32:27.399042+00:00",{"id":63,"slug":64,"title":65,"cover_image":66,"image_url":66,"created_at":67,"category":13},"edc921e7-46eb-457f-b063-c69ca74bce98","agent-skills-llm-agents-next-layer-zh","技能層：LLM Agent 下一層","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784620982310-7v8r.png","2026-07-21T08:02:29.196519+00:00",{"id":69,"slug":70,"title":71,"cover_image":72,"image_url":72,"created_at":73,"category":13},"cc2c9df3-f18b-4c01-b61e-84f46296c0e5","offline-first-llm-low-connectivity-learning-zh","離線優先 LLM，救低網速學習","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784619184562-aw7y.png","2026-07-21T07:32:28.395731+00:00",[75,80,85,90,95,100,105,110,115,120],{"id":76,"slug":77,"title":78,"created_at":79},"f18dbadb-8c59-4723-84a4-6ad22746c77a","deepmind-bets-on-continuous-learning-ai-2026-zh","DeepMind 押注 2026 連續學習 AI","2026-03-26T08:16:02.367355+00:00",{"id":81,"slug":82,"title":83,"created_at":84},"f4a106cb-02a6-4508-8f39-9720a0a93cee","ml-papers-of-the-week-github-research-desk-zh","每週 ML 論文清單，為何紅到 GitHub","2026-03-27T01:11:39.284175+00:00",{"id":86,"slug":87,"title":88,"created_at":89},"c4f807ca-4e5f-47f1-a48c-961cf3fc44dc","ai-ml-conferences-to-watch-in-2026-zh","2026 AI 研討會投稿時程整理","2026-03-27T01:51:53.874432+00:00",{"id":91,"slug":92,"title":93,"created_at":94},"cf046742-efb2-4753-aef9-caed5da5e32e","adaptive-block-scaled-data-types-zh","IF4：神經網路量化的聰明選擇","2026-03-31T06:00:36.990273+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"53a0dc54-0371-4e40-8d5e-74e94a73840c","geometry-aware-similarity-metrics-for-neural-representations-zh","超越距離測量：用微分幾何重新理解神經網路","2026-03-31T06:01:01.241968+00:00",{"id":101,"slug":102,"title":103,"created_at":104},"fee7d472-a775-4b1d-bbc2-1e8bca1bbf8b","on-the-fly-repulsion-in-the-contextual-space-for-rich-divers-zh","讓AI繪圖更有創意：用排斥力提升生成多樣性","2026-03-31T06:01:25.439673+00:00",{"id":106,"slug":107,"title":108,"created_at":109},"a9901203-d69b-447b-8854-15d14eab32b4","vision-aided-beam-prediction-cnn-eca-zh","影像輔助波束預測升級 CNN","2026-04-01T10:00:25.8073+00:00",{"id":111,"slug":112,"title":113,"created_at":114},"b55e7dd4-0a24-4b3d-804d-b0309a03f498","triple-band-fss-mimo-antenna-sub-6-ghz-zh","三頻 FSS MIMO 天線瞄準 sub-6 GHz","2026-04-01T13:18:36.857305+00:00",{"id":116,"slug":117,"title":118,"created_at":119},"f68290bd-e7f3-4b30-ba22-dcd4e0130a66","openclaw-1299-repos-eight-weeks-analysis-zh","OpenClaw 1299 個 Repo 的資料解讀","2026-04-02T05:03:45.208411+00:00",{"id":121,"slug":122,"title":123,"created_at":124},"ed9f80eb-eb02-4d35-8ad4-0ddf428751dd","beam-coherence-aware-combining-mmwave-mimo-zh","毫米波 MIMO 的雙階合併法","2026-04-02T05:27:26.897188+00:00"]