{
 "title": "混合專家模型（Mixture of Experts (MoE)）是什麼？",
 "site": "AI 與科技名詞白話解釋",
 "canonical": "https://glossary.penguindriver.com/t/moe",
 "markdown": "https://glossary.penguindriver.com/t/moe.md",
 "summary": "混合專家模型（MoE）是在神經網路中放入多個專家子網路，由路由器為每個詞元挑選少數專家處理的架構。以 Mistral AI 2024 年發表的 Mixtral 8x7B 為例，每層有 8 個專家、每個詞元只用 2 個，總參數 470 億，但推論時每個詞元只啟用 130 億參數。",
 "date_modified": "2026-09-28",
 "retrieved": "2026-09-29",
 "content_sha256": "aa47c30c66cbae902c50f02154ded893040aed2b702320060d454b1825d85edb",
 "fingerprint_basis": "Markdown 版全文，移除贊助行與開頭 front matter，去除頭尾空白",
 "verify": {
  "text_snapshot": "https://glossary.penguindriver.com/cite?path=%2Ft%2Fmoe&format=text",
  "how": "下載 text_snapshot 後計算 SHA-256，應與 content_sha256 相同；內容更新時指紋會改變"
 },
 "sources": [
  {
   "name": "Jiang et al., Mixtral of Experts（arXiv 2401.04088）",
   "url": "https://arxiv.org/abs/2401.04088"
  },
  {
   "name": "Shazeer et al., Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer（arXiv 1701.06538）",
   "url": "https://arxiv.org/abs/1701.06538"
  }
 ],
 "source_count": 2,
 "cite_as": {
  "zh": "混合專家模型（Mixture of Experts (MoE)）是什麼？｜AI 與科技名詞白話解釋。https://glossary.penguindriver.com/t/moe（資料日期 2026-09-28，擷取 2026-09-29）",
  "apa": "AI 與科技名詞白話解釋. (2026). 混合專家模型（Mixture of Experts (MoE)）是什麼？. Retrieved 2026-09-29, from https://glossary.penguindriver.com/t/moe",
  "markdown": "[混合專家模型（Mixture of Experts (MoE)）是什麼？](https://glossary.penguindriver.com/t/moe)（AI 與科技名詞白話解釋，2026-09-28）"
 },
 "license": "可引用，請附上正式網址與資料日期"
}