[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-leanstral-1-5-open-source-math-models-useful-zh":3,"article-related-leanstral-1-5-open-source-math-models-useful-zh":31,"series-research-be6cb41d-ace9-4841-b335-2d017de565cc":75},{"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},"be6cb41d-ace9-4841-b335-2d017de565cc","leanstral-1-5-open-source-math-models-useful-zh","Leanstral 1.5 證明開源數學模型已經能上場","\u003Cp data-speakable=\"summary\">100% 的 miniF2F 成績證明，開源 Lean \u003Ca href=\"\u002Fnews\u002Fdirect-opd-weak-to-strong-distillation-zh\">模型\u003C\u002Fa>已經能勝任形式化證明工作。\u003C\u002Fp>\u003Cp>Leanstral 1.5 不是又一個會聊天的模型，而是一個能在形式化數學與程式驗證上交付成果的開源系統；它的 \u003Ca href=\"\u002Ftag\u002Fbenchmark\">benchmark\u003C\u002Fa> 表現已經明確越過「看起來厲害」的門檻，進入「真的有用」的區間。\u003C\u002Fp>\u003Ch2>第一個論點\u003C\u002Fh2>\u003Cp>在 miniF2F 上拿到 100% 並不是漂亮數字而已。miniF2F 涵蓋從高中數學到奧林匹亞難度的形式化題目，這代表模型不是在背題庫，而是在 Lean 4 裡穩定處理結構化證明。當一個系統能把這類任務做滿分，說明它已經掌握了形式推理的核心操作。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783449168809-rp22.png\" alt=\"Leanstral 1.5 證明開源數學模型已經能上場\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>更難忽視的是 PutnamBench：672 題裡解出 587 題，這不是邊角料成績。Mistral 也指出，它在 PutnamBench、FATE-H、FATE-X 都領先開源陣營；只有閉源的 Aleph Prover 在 PutnamBench 上更強。這個差距很小，卻足以說明一件事，\u003Ca href=\"\u002Ftag\u002F開源模型\">開源模型\u003C\u002Fa>已經逼近最頂尖的商用系統。\u003C\u002Fp>\u003Ch2>第二個論點\u003C\u002Fh2>\u003Cp>形式化數學比一般聊天 benchmark 更嚴格，因為它不吃語氣、不吃流暢度，只看證明能不能編譯。這種評測方式把「像是在思考」和「真的推導出正確結果」清楚分開，所以 Leanstral 1.5 的表現具有更高可信度，也更接近工程世界需要的可靠性。\u003C\u002Fp>\u003Cp>FATE-H 和 FATE-X 的結果進一步補強這點。這兩個 benchmark 涵蓋碩士與博士層級的代數題，包含群論與環論，而 Leanstral 1.5 分別拿到 87% 與 34% 的開源最佳成績。這表示它不是只會某一類題型，而是能在一整族依賴符號精確性的任務上維持能力。\u003C\u002Fp>\u003Ch2>反方可能怎麼說\u003C\u002Fh2>\u003Cp>最強的反對意見很直接：benchmark 再高，也不等於日常開發真的有用。Lean 4 是高度專門的環境，形式化證明的世界和一般軟體工程差很多；模型在定理證明裡表現好，不代表它能理解模糊需求、混亂 codebase，或處理產品決策中的灰色地帶。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783449160334-tf7r.png\" alt=\"Leanstral 1.5 證明開源數學模型已經能上場\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這個質疑成立，而且它應該被保留。Leanstral 1.5 的價值不在於宣稱自己能取代所有工程工作，而在於它已經在最可驗證的場景裡證明了自己：它不只在 benchmark 上強，還真的掃過 57 個開源 repo，找出 5 個先前未知的 bug，其中包含 \u003Ca href=\"\u002Ftag\u002Frust\">Rust\u003C\u002Fa> 函式庫 varinteger 的 overflow 問題。這不是抽象能力，而是可落地的修補價值。\u003C\u002Fp>\u003Ch2>你能做什麼\u003C\u002Fh2>\u003Cp>工程師應把 Leanstral 1.5 當成驗證助手，用在 proof sketch、invariant 檢查、關鍵路徑 bug hunting，不要把它當萬能聊天\u003Ca href=\"\u002Fnews\u002Fcamvla-calibration-free-view-robust-vla-zh\">機器人\u003C\u002Fa>；PM 應優先在編譯器、資安、金融、\u003Ca href=\"\u002Fnews\u002Friyadh-blockchain-show-web3-infrastructure-zh\">基礎設施\u003C\u002Fa>這類錯誤成本高的場景試點；創辦人則該把訊號看清楚，開源模型已經足以切入「信任建立在形式證明」的領域，下一批真正耐久的 AI 產品，很可能就長在這裡。\u003C\u002Fp>","Leanstral 1.5 的成績顯示，開源模型已經能在形式化數學與程式驗證上提供實際價值，不再只是展示型作品。","the-decoder.com","https:\u002F\u002Fthe-decoder.com\u002Fmistrals-open-source-leanstral-1-5-aces-formal-math-benchmarks-and-catches-real-bugs-in-code\u002F",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783449168809-rp22.png","research","zh","ed4393a9-437c-4266-b94c-b51ed66acc7b",[17,18,19,20,21,22],"Leanstral 1.5","開源模型","形式化數學","Lean 4","程式驗證","bug 發現",[24,25,26],"miniF2F 100% 與 PutnamBench 成績，證明開源模型已能處理高難度形式化證明。","形式化數學比聊天 benchmark 更能測出真正推理能力，因此結果可信度更高。","實際掃描 repo 找到未知 bug，顯示這類模型已能直接產生工程價值。",3,"2026-07-07T18:32:18.210433+00:00","2026-07-07T18:32:18.201+00:00","be014adf-10f8-451b-831d-952092bb9d8c",{"tags":32,"relatedLang":34,"relatedPosts":38},[33],{"name":18,"slug":18},{"id":15,"slug":35,"title":36,"language":37},"leanstral-1-5-open-source-math-models-useful-en","Leanstral 1.5 proves open-source math models are now useful","en",[39,45,51,57,63,69],{"id":40,"slug":41,"title":42,"cover_image":43,"image_url":43,"created_at":44,"category":13},"a1c5b218-d9ff-4e46-9c58-07d0fe5152fc","vlm-accuracy-visual-cognitive-errors-decade-zh","VLM 描述複雜場景變準了","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783926189859-c95z.png","2026-07-13T07:02:36.585294+00:00",{"id":46,"slug":47,"title":48,"cover_image":49,"image_url":49,"created_at":50,"category":13},"2ec5f4bf-f90a-4dc9-98e0-dc8189169e56","visual-pretraining-language-models-zh","視覺預訓練勝過純文字","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783924384413-4ob9.png","2026-07-13T06:32:35.520894+00:00",{"id":52,"slug":53,"title":54,"cover_image":55,"image_url":55,"created_at":56,"category":13},"8b8f7b87-7e93-415f-a52d-56613e17b278","phinn-eeg-topology-dream-state-eeg-zh","PHINN-EEG 用拓撲看夢境 EEG","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783922588253-kq48.png","2026-07-13T06:02:34.287269+00:00",{"id":58,"slug":59,"title":60,"cover_image":61,"image_url":61,"created_at":62,"category":13},"c4597538-217d-4b81-83d0-9b3cc4153861","google-android-bench-update-gemini-gap-zh","Android Bench 更新，Gemini 掉到第五","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783906366388-1v3j.png","2026-07-13T01:32:25.247653+00:00",{"id":64,"slug":65,"title":66,"cover_image":67,"image_url":67,"created_at":68,"category":13},"f25ed4f5-db61-4d8c-bc59-e80c93e27927","llm-benchmarks-not-enough-2026-zh","2026 年挑 LLM，別再把 benchmark 當答案","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783818161840-t0n4.png","2026-07-12T01:02:19.419242+00:00",{"id":70,"slug":71,"title":72,"cover_image":73,"image_url":73,"created_at":74,"category":13},"35378c9f-bc39-4cc0-b9e1-1ce4a746ba5b","rust-breaks-into-tiobe-top-10-zh","Rust 進入 TIOBE 前十的判讀筆記","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783816365723-zjl1.png","2026-07-12T00:32:23.969578+00:00",[76,81,86,91,96,101,106,111,116,121],{"id":77,"slug":78,"title":79,"created_at":80},"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":82,"slug":83,"title":84,"created_at":85},"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":87,"slug":88,"title":89,"created_at":90},"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":92,"slug":93,"title":94,"created_at":95},"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":97,"slug":98,"title":99,"created_at":100},"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":102,"slug":103,"title":104,"created_at":105},"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":107,"slug":108,"title":109,"created_at":110},"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":112,"slug":113,"title":114,"created_at":115},"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":117,"slug":118,"title":119,"created_at":120},"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":122,"slug":123,"title":124,"created_at":125},"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"]