[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-fine-tuning":3},{"tag":4,"articles":10,"peer_article_count":79},{"id":5,"name":6,"slug":6,"article_count":7,"description_zh":8,"description_en":9},"3725f3eb-5764-4f52-a203-bdfaacae8acc","fine-tuning",4,"微調是把通用模型改造成特定任務工具的關鍵步驟，常見於新詞注入、指令對齊與多模態適配。重點不只在訓練技巧，也在初始化、資料分佈、VRAM 需求與語言覆蓋，直接影響生成品質與部署成本。","Fine-tuning adapts a base model to a narrower task or domain, from seeding new vocabulary and aligning instruction behavior to adapting vision-language models. The practical issues are initialization, data quality, VRAM limits, and language coverage, all of which shape output quality and deployment cost.",[11,20,28,36,43,50,57,64,72],{"id":12,"slug":13,"title":14,"summary":15,"category":16,"image_url":17,"cover_image":17,"language":18,"created_at":19},"c40b20df-d89a-43ae-bb11-11062dcd2cd2","llms-work-by-predicting-next-token-zh","5 個關鍵部件看懂 LLMs","5 個關鍵部件帶你看懂 LLMs 如何從資料、token、注意力到對齊，進而判斷訓練與部署該看什麼。","industry","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781889466449-7e9g.png","zh","2026-06-19T17:17:20.910277+00:00",{"id":21,"slug":22,"title":23,"summary":24,"category":25,"image_url":26,"cover_image":26,"language":18,"created_at":27},"09e34016-bbc0-4313-b090-2dbfdd6cf96a","fine-tuning-slms-turns-enterprise-ai-practical-zh","SLM 微調把企業 AI 變可用","拆 CogitX 的 SLM 微調 playbook，整理成企業訓練、評估、部署都能直接照抄的模板。","ai-agent","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781359406320-5jrq.png","2026-06-13T14:02:55.242488+00:00",{"id":29,"slug":30,"title":31,"summary":32,"category":33,"image_url":34,"cover_image":34,"language":18,"created_at":35},"85a3dfd4-9f87-4028-8ec1-44636ec804b8","why-fine-tuning-still-beats-prompt-only-ai-zh","為什麼微調仍然勝過只靠提示詞的 AI","微調仍是把基礎模型做成可靠專用工具的最佳方法，因為它改變模型本身，而不只是包裝在外的提示詞。","research","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780120970410-mtab.png","2026-05-30T06:02:21.657693+00:00",{"id":37,"slug":38,"title":39,"summary":40,"category":33,"image_url":41,"cover_image":41,"language":18,"created_at":42},"a9b25f93-fd42-4aa7-95be-c4e648ad48c7","how-to-build-ai-research-foundations-with-deepmind-zh","怎麼用 DeepMind 建立 AI 研究基礎","這篇教你用 DeepMind 的課程與本機實作，建立現代語言模型與微調的入門基礎。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779963521409-34ys.png","2026-05-28T10:17:24.309519+00:00",{"id":44,"slug":45,"title":46,"summary":47,"category":33,"image_url":48,"cover_image":48,"language":18,"created_at":49},"d1c6850c-f832-471b-8beb-c0ebc809667d","peft-bench-fine-tuning-methods-benchmark-zh","PEFT-Bench 讓微調比較更公平","PEFT-Bench 把 27 個 NLP 資料集與 7 種 PEFT 方法放進同一套流程，比的不只準確率，也把參數、速度和記憶體成本算進去。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779179048497-jm5y.png","2026-05-19T08:23:36.803043+00:00",{"id":51,"slug":52,"title":53,"summary":54,"category":33,"image_url":55,"cover_image":55,"language":18,"created_at":56},"f659bb6c-0788-4653-a1b6-53f8798c8564","microsoft-goalcover-fine-tuning-gaps-zh","Microsoft GoalCover 找出微調缺口","Microsoft Research 的 GoalCover 會在微調前找出資料缺口，並在 Qwen-3-14B 的金融摘要任務上提升 reward 分數。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778462447499-4gq4.png","2026-05-11T01:20:33.736263+00:00",{"id":58,"slug":59,"title":60,"summary":61,"category":33,"image_url":62,"cover_image":62,"language":18,"created_at":63},"72828ff9-cbfb-4e10-81b2-9c4c9544b7f1","how-to-build-vintage-llm-testbed-5-steps-zh","5 步建出 1930 截止 LLM 測試台","用 5 個步驟建立一個 1930 截止的 LLM 測試台，驗證歷史推理與無污染泛化。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777945255070-tzh6.png","2026-05-05T01:40:31.318037+00:00",{"id":65,"slug":66,"title":67,"summary":68,"category":69,"image_url":70,"cover_image":70,"language":18,"created_at":71},"1d6cf1d5-8fc4-41f5-9802-68b115639cee","grounded-token-initialization-new-vocabulary-zh","新詞初始化別再只取平均","GTI 在微調前先把新詞錨定到有意義的嵌入位置，避免平均初始化讓 token 擠成一團，進而影響生成式推薦的表現。","blockchain","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775196588195-1xuy.png","2026-04-03T06:09:29.631723+00:00",{"id":73,"slug":74,"title":75,"summary":76,"category":16,"image_url":77,"cover_image":77,"language":18,"created_at":78},"fa006110-18df-40ed-ac43-1e2133fa2c06","rag-in-2026-what-enterprise-ai-needs-now-zh","2026 年企業 AI 為何更靠 RAG","RAG 已從展示用技術走進企業預算。原因很直接：公司要的是能讀取最新內部資料、可追溯、可控權限的 AI，而不是只會背舊訓練資料的聊天模型。到了 2026 年，真正有用的重點在檢索品質、權限治理、即時資料連接與合規設計。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1774517341124-yymy.png","2026-03-26T08:06:06.808873+00:00",14]