{"id":2519,"date":"2025-06-21T08:39:55","date_gmt":"2025-06-21T00:39:55","guid":{"rendered":"https:\/\/sanlangcode.com\/?p=2519"},"modified":"2025-06-21T08:53:45","modified_gmt":"2025-06-21T00:53:45","slug":"%e4%bd%bf%e7%94%a8qwen3%e7%9a%84reranker%e6%a8%a1%e5%9e%8bembedding%e5%ae%9e%e7%8e%b0%e6%9f%a5%e6%89%be%e4%b8%8e%e7%b2%be%e6%8e%92","status":"publish","type":"post","link":"https:\/\/sanlangcode.com\/index.php\/2025\/06\/21\/%e4%bd%bf%e7%94%a8qwen3%e7%9a%84reranker%e6%a8%a1%e5%9e%8bembedding%e5%ae%9e%e7%8e%b0%e6%9f%a5%e6%89%be%e4%b8%8e%e7%b2%be%e6%8e%92\/","title":{"rendered":"\u4f7f\u7528Qwen3\u7684Embedding\u548cReranker\u6a21\u578b\u5b9e\u73b0\u67e5\u627e\u4e0e\u7cbe\u6392"},"content":{"rendered":"\n

\u6700\u8fd1\u51e0\u5929\u963f\u91cc\u53d1\u5e03\u4e86Reranker\u6a21\u578bEmbedding\uff0c\u5c1d\u8bd5\u4f7f\u7528\u5b83\u4e24\u4e2a\u6765\u7cbe\u51c6\u67e5\u627e\u68c0\u7d22\u6587\u4ef6\u5185\u5bb9\u3002<\/p>\n\n\n\n

\u9879\u76ee\u5730\u5740\uff1aSanroZhang\/qwen3-ER-search (github.com)<\/a><\/p>\n\n\n\n

\ud83d\udd0d Reranker vs Embedding \u6a21\u578b\u5bf9\u6bd4<\/h2>\n\n\n\n

Embedding\u6a21\u578b<\/strong> (Qwen3-Embedding)<\/h3>\n\n\n\n
    \n
  • \u529f\u80fd<\/strong>: \u5c06\u6587\u672c\u8f6c\u6362\u4e3a\u5411\u91cf\uff0c\u8ba1\u7b97\u8bed\u4e49\u76f8\u4f3c\u5ea6<\/li>\n\n\n\n
  • \u7528\u9014<\/strong>: \u7c97\u7b5b\u9636\u6bb5\uff0c\u4ece\u5927\u91cf\u6587\u6863\u4e2d\u5feb\u901f\u627e\u5230\u76f8\u5173\u5019\u9009<\/li>\n\n\n\n
  • \u7279\u70b9<\/strong>: \u901f\u5ea6\u5feb\uff0c\u4f46\u7cbe\u5ea6\u53ef\u80fd\u4e0d\u591f\u9ad8<\/li>\n<\/ul>\n\n\n\n

    Reranker\u6a21\u578b<\/strong> (Qwen3-Reranker)<\/h3>\n\n\n\n
      \n
    • \u529f\u80fd<\/strong>: \u5bf9\u5019\u9009\u6587\u6863\u8fdb\u884c\u7cbe\u786e\u91cd\u6392\u5e8f\uff0c\u5224\u65ad\u76f8\u5173\u6027<\/li>\n\n\n\n
    • \u7528\u9014<\/strong>: \u7cbe\u7b5b\u9636\u6bb5\uff0c\u5bf9Embedding\u7684\u7ed3\u679c\u8fdb\u884c\u4e8c\u6b21\u6392\u5e8f<\/li>\n\n\n\n
    • \u7279\u70b9<\/strong>: \u7cbe\u5ea6\u9ad8\uff0c\u4f46\u901f\u5ea6\u76f8\u5bf9\u8f83\u6162<\/li>\n<\/ul>\n\n\n\n

      \ud83c\udfaf \u4e24\u9636\u6bb5\u68c0\u7d22\u6d41\u7a0b<\/h2>\n\n\n\n
      \u6587\u6863\u5e93 \u2192 Embedding\u7c97\u7b5b \u2192 \u5019\u9009\u6587\u6863 \u2192 Reranker\u7cbe\u7b5b \u2192 \u6700\u7ec8\u7ed3\u679c<\/code><\/pre>\n\n\n\n
      \u521b\u5efa\u4e00\u4e2a\u7ed3\u5408\u4e24\u79cd\u6a21\u578b\u7684\u7cbe\u51c6\u68c0\u7d22\u7cfb\u7edf\uff1a<\/p>\n\n\n\n
      \u8fd0\u884c\u6df7\u5408\u68c0\u7d22\u7cfb\u7edf\uff0c\u5c55\u793aEmbedding\u548cReranker\u7684\u5bf9\u6bd4\u6548\u679c\uff1a<\/p>\n\n\n\n
      python hybrid_retrieval.py\n\n\u7ed3\u679c\uff1a\n\n<\/code><\/pre>\n\n\n\n
      PS C:\\Users\\sanrome\\Documents\\zero_nlp-main\\qwen3_embedding> python hybrid_retrieval.py\n\u6b63\u5728\u52a0\u8f7dQwen3 Embedding\u6a21\u578b...\nEmbedding\u6a21\u578b\u52a0\u8f7d\u5b8c\u6210\uff01\nReranker\u6a21\u578b\u52a0\u8f7d\u5b8c\u6210\uff01\n\u6b63\u5728\u52a0\u8f7dPDF\u6587\u4ef6: C:\\Users\\sanrome\\Documents\\\u4e09\u90ce\u767d\u5e95\u901a\u7528\u7b80\u5386-zh.pdf\n\u6210\u529f\u63d0\u53d6 12 \u4e2a\u6587\u672c\u5757\n\u6b63\u5728\u6784\u5efa\u6587\u6863\u5411\u91cf...\n\u6587\u6863\u5411\u91cf\u6784\u5efa\u5b8c\u6210\n\n================================================================================\n=== \u6df7\u5408\u68c0\u7d22\u7cfb\u7edf\u6d4b\u8bd5 ===\n================================================================================\n\n============================================================\n\u67e5\u8be2: \u674e\u660e\u7684\u5de5\u4f5c\u7ecf\u9a8c\n============================================================\n\n1\ufe0f\u20e3 \u4ec5Embedding\u641c\u7d22:\n\u4ec5\u4f7f\u7528Embedding\u641c\u7d22: \u674e\u660e\u7684\u5de5\u4f5c\u7ecf\u9a8c\n  \u7ed3\u679c1: \u76f8\u4f3c\u5ea6=0.4475, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406\n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317\u534e\u5927\u5b66-\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u4e13\u4e1a\u3002\u6211\u76ee\u524d\u5df2\u7ecf\n...\n  \u7ed3\u679c2: \u76f8\u4f3c\u5ea6=0.3059, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a\u6d4b\u8bd5Bug \u5e76\u8bb0\u5f55\u590d\u73b0\u8fc7\u7a0b\u548c\u6d4b\u8bd5\u7ed3\u679c \uff0c\u5e2e...\n  \u7ed3\u679c3: \u76f8\u4f3c\u5ea6=0.2979, \u7d22\u5f15=1\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u56fe\u50cf\u5904\u7406\u4e0e\u6807\u6ce8\u8f6f\u4ef6\u5957\u88c5\u7684\u5f00\u53d1\u4e0e\u6d4b\u8bd5\u3002\u5bf9\u63a5\u5ba2\u6237\u9700\u6c42\uff0c\u8d1f\u8d23\u8f6f\u4ef6\u67b6\u6784\u8bbe\u3002\u8ba1\uff0c\u8bbe\u8ba1UI \u539f\u578b\uff0c \n\u8c03\u7814\u76f8\u540c\u7ade\u54c1\u5f62\u6210\u62a5\u544a\uff0c\u786e\u5b9a\u8f6f\u4ef6\u6280\u672f\u67b6\u6784\u548c\u6a21\u5757\u529f\u80fd\u5212\u5206\u3002\u4e3b\u3002\u5bfc\u63a7\u5236\u6a21\u5757\uff0c\u6839\u636e\u5de5\u4e1a\u901a\u8baf\u534f\u8bae\u5bf9\u63a5\u4e0b\u4f4d\u673a\u5b9e \u73b0\u8bbe\u5907\u9a71\u52a8...\n                                             \n2\ufe0f\u20e3 \u4ec5Reranker\u641c\u7d22:\n\u4ec5\u4f7f\u7528Reranker\u641c\u7d22: \u674e\u660e\u7684\u5de5\u4f5c\u7ecf\u9a8c\nYou're using a Qwen2TokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__c\nall__` method is faster than using a method to encode the text followed by a call to the `pad` meth\nod to get a padded encoding.\nC:\\Users\\sanrome\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\site-packages\\transformers\\tokenizatio\nn_utils_base.py:2714: UserWarning: `max_length` is ignored when `padding`=`True` and there is no tr\nuncation strategy. To pad to max length, use `padding='max_length'`.\n  warnings.warn(\n  \u7ed3\u679c1: \u76f8\u5173\u6027=0.9995, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317\u534e\u5927\u5b66-\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u4e13\u4e1a\u3002\u6211\u76ee\u524d\u5df2\u7ecf...\n                                                      \u7ed3\u679c2: \u76f8\u5173\u6027=0.9819, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002-\u3002\u5317\u4eac\u30024. \u8da3\u5934\u6761- \u6d4b\u8bd5\u5b9e\u4e60\u751f\u3002-...                         \u7ed3\u679c3: \u76f8\u5173\u6027=0.0158, \u7d22\u5f15=11\n    \u5185\u5bb9: 5. CSDN\/\u963f\u91cc\u4e91\u5f00\u53d1\u8005\u3002\u793e\u533a\u6280\u672f\u535a\u4e3b\u30026. \u963f\u91cc\u4e91\u670d\u52a1\u5668\u8fdb\u9636\u5b9e\u3002\u6218\u8bad\u7ec3\u8425\u30027. \u4e2d\u6d77\u56fd\u9645\u6821\u56ed\u4f01\u4e1a \n\u5b9e\u3002\u8bad\u3002-\u5c0f\u7ec4\u603b\u5206\u7b2c1 \u4f18\u80dc\u30028. \u5927\u5b66\u751f\u201c\u6311\u6218\u676f\u201d\u521b\u3002\u65b0\u521b\u4e1a\u5927\u8d5b-\u6821\u7ea7\u4e8c\u7b49\u3002\u5956\u3001\u7701\u7ea7\u4e09\u7b49\u5956\u30029....         \n3\ufe0f\u20e3 \u6df7\u5408\u641c\u7d22 (Embedding + Reranker):\n\u6b63\u5728\u8fdb\u884c\u6df7\u5408\u68c0\u7d22: \u674e\u660e\u7684\u5de5\u4f5c\u7ecf\u9a8c\n==================================================\n\u7b2c\u4e00\u9636\u6bb5\uff1aEmbedding\u7c97\u7b5b\nEmbedding\u9636\u6bb5\u627e\u5230 5 \u4e2a\u5019\u9009\u6587\u6863:\n  \u5019\u90091: \u76f8\u4f3c\u5ea6=0.4475, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317...                                      \u5019\u90092: \u76f8\u4f3c\u5ea6=0.3059, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a...                                \u5019\u90093: \u76f8\u4f3c\u5ea6=0.2979, \u7d22\u5f15=1\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u56fe\u50cf\u5904\u7406\u4e0e\u6807\u6ce8\u8f6f\u4ef6\u5957\u88c5\u7684\u5f00\u53d1\u4e0e\u6d4b\u8bd5\u3002\u5bf9\u63a5\u5ba2\u6237\u9700\u6c42\uff0c\u8d1f\u8d23\u8f6f\u4ef6\u67b6\u6784\u8bbe\u3002\u8ba1\uff0c\u8bbe\u8ba1UI \u539f\u578b\uff0c \n\u8c03\u7814\u76f8\u540c\u7ade\u54c1\u5f62\u6210\u62a5\u544a\uff0c\u786e\u5b9a\u8f6f\u4ef6\u6280\u672f\u67b6\u6784\u548c\u6a21\u5757\u529f\u80fd\u5212\u5206\u3002\u4e3b\u3002\u5bfc\u63a7\u5236\u6a21\u5757...                              \u5019\u90094: \u76f8\u4f3c\u5ea6=0.2690, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002...                                                          \u5019\u90095: \u76f8\u4f3c\u5ea6=0.2444\n, \u7d22\u5f15=3\n    \u5185\u5bb9: 4. AI \u5927\u6a21\u578b\uff1a\u638c\u63e1DeepSeek \u7b49\u5927\u6a21\u578b\u90e8\u7f72\u4e0e\u8c03\u4f18\uff0c\u719f\u6089\u534f\u8bae\u5b9a\u4e49\u4e0e\u63a5\u53e3\u5f00\u53d1\u4f7f\u7528\u30025. \u6d4b\u8bd5\u6280\u80fd\uff1a \n\u719f\u6089\u5355\u5143\u6d4b\u8bd5\u3001\u63a5\u53e3\u6d4b\u8bd5\u3001UI \u6d4b\u8bd5\u7684\u6279\u5904\u7406\u4e0e\u81ea\u52a8\u5316\u811a\u672c...                                             \n\u7b2c\u4e8c\u9636\u6bb5\uff1aReranker\u7cbe\u7b5b\nReranker\u9636\u6bb5\u91cd\u65b0\u6392\u5e8f\u7ed3\u679c:\n  \u6392\u5e8f1: Reranker\u5206\u6570=0.9995, Embedding\u5206\u6570=0.4475, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317...                                      \u6392\u5e8f2: Reranker\u5206\u6570=0.9819, Em\nbedding\u5206\u6570=0.2690, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002...                                                          \u6392\u5e8f3: Reranker\u5206\u6570\n=0.0028, Embedding\u5206\u6570=0.2979, \u7d22\u5f15=1\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u56fe\u50cf\u5904\u7406\u4e0e\u6807\u6ce8\u8f6f\u4ef6\u5957\u88c5\u7684\u5f00\u53d1\u4e0e\u6d4b\u8bd5\u3002\u5bf9\u63a5\u5ba2\u6237\u9700\u6c42\uff0c\u8d1f\u8d23\u8f6f\u4ef6\u67b6\u6784\u8bbe\u3002\u8ba1\uff0c\u8bbe\u8ba1UI \u539f\u578b\uff0c \n\u8c03\u7814\u76f8\u540c\u7ade\u54c1\u5f62\u6210\u62a5\u544a\uff0c\u786e\u5b9a\u8f6f\u4ef6\u6280\u672f\u67b6\u6784\u548c\u6a21\u5757\u529f\u80fd\u5212\u5206\u3002\u4e3b\u3002\u5bfc\u63a7\u5236\u6a21\u5757...                              \u6392\u5e8f4: Reranker\u5206\u6570=0.0008, Embeddin\ng\u5206\u6570=0.3059, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a...                                \u6392\u5e8f5: Reranker\u5206\u6570=0.0002, Embed\nding\u5206\u6570=0.2444, \u7d22\u5f15=3\n    \u5185\u5bb9: 4. AI \u5927\u6a21\u578b\uff1a\u638c\u63e1DeepSeek \u7b49\u5927\u6a21\u578b\u90e8\u7f72\u4e0e\u8c03\u4f18\uff0c\u719f\u6089\u534f\u8bae\u5b9a\u4e49\u4e0e\u63a5\u53e3\u5f00\u53d1\u4f7f\u7528\u30025. \u6d4b\u8bd5\u6280\u80fd\uff1a \n\u719f\u6089\u5355\u5143\u6d4b\u8bd5\u3001\u63a5\u53e3\u6d4b\u8bd5\u3001UI \u6d4b\u8bd5\u7684\u6279\u5904\u7406\u4e0e\u81ea\u52a8\u5316\u811a\u672c...                                               \u7ed3\u679c1: \u6700\u7ec8\u76f8\u5173\u6027=0.9995, \u7d22\u5f15=\n0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317\u534e\u5927\u5b66-\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u4e13\u4e1a\u3002\u6211\u76ee\u524d\u5df2\u7ecf...\n                                                      \u7ed3\u679c2: \u6700\u7ec8\u76f8\u5173\u6027=0.9819, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002-\u3002\u5317\u4eac\u30024. \u8da3\u5934\u6761- \u6d4b\u8bd5\u5b9e\u4e60\u751f\u3002-...                         \u7ed3\u679c3: \u6700\u7ec8\u76f8\u5173\u6027=0.0028, \u7d22\u5f15=1\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u56fe\u50cf\u5904\u7406\u4e0e\u6807\u6ce8\u8f6f\u4ef6\u5957\u88c5\u7684\u5f00\u53d1\u4e0e\u6d4b\u8bd5\u3002\u5bf9\u63a5\u5ba2\u6237\u9700\u6c42\uff0c\u8d1f\u8d23\u8f6f\u4ef6\u67b6\u6784\u8bbe\u3002\u8ba1\uff0c\u8bbe\u8ba1UI \u539f\u578b\uff0c \n\u8c03\u7814\u76f8\u540c\u7ade\u54c1\u5f62\u6210\u62a5\u544a\uff0c\u786e\u5b9a\u8f6f\u4ef6\u6280\u672f\u67b6\u6784\u548c\u6a21\u5757\u529f\u80fd\u5212\u5206\u3002\u4e3b\u3002\u5bfc\u63a7\u5236\u6a21\u5757\uff0c\u6839\u636e\u5de5\u4e1a\u901a\u8baf\u534f\u8bae\u5bf9\u63a5\u4e0b\u4f4d\u673a\u5b9e \u73b0\u8bbe\u5907\u9a71\u52a8...\n                                             \n============================================================\n\u67e5\u8be2: \u6280\u672f\u6280\u80fd\u548c\u7f16\u7a0b\u8bed\u8a00\n============================================================\n\n1\ufe0f\u20e3 \u4ec5Embedding\u641c\u7d22:\n\u4ec5\u4f7f\u7528Embedding\u641c\u7d22: \u6280\u672f\u6280\u80fd\u548c\u7f16\u7a0b\u8bed\u8a00\n  \u7ed3\u679c1: \u76f8\u4f3c\u5ea6=0.3679, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317\u534e\u5927\u5b66-\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u4e13\u4e1a\u3002\u6211\u76ee\u524d\u5df2\u7ecf...\n                                                      \u7ed3\u679c2: \u76f8\u4f3c\u5ea6=0.3538, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a\u6d4b\u8bd5Bug \u5e76\u8bb0\u5f55\u590d\u73b0\u8fc7\u7a0b\u548c\u6d4b\u8bd5\u7ed3\u679c \uff0c\u5e2e...\n                                                      \u7ed3\u679c3: \u76f8\u4f3c\u5ea6=0.3130, \u7d22\u5f15=11\n    \u5185\u5bb9: 5. CSDN\/\u963f\u91cc\u4e91\u5f00\u53d1\u8005\u3002\u793e\u533a\u6280\u672f\u535a\u4e3b\u30026. \u963f\u91cc\u4e91\u670d\u52a1\u5668\u8fdb\u9636\u5b9e\u3002\u6218\u8bad\u7ec3\u8425\u30027. \u4e2d\u6d77\u56fd\u9645\u6821\u56ed\u4f01\u4e1a \n\u5b9e\u3002\u8bad\u3002-\u5c0f\u7ec4\u603b\u5206\u7b2c1 \u4f18\u80dc\u30028. \u5927\u5b66\u751f\u201c\u6311\u6218\u676f\u201d\u521b\u3002\u65b0\u521b\u4e1a\u5927\u8d5b-\u6821\u7ea7\u4e8c\u7b49\u3002\u5956\u3001\u7701\u7ea7\u4e09\u7b49\u5956\u30029....         \n2\ufe0f\u20e3 \u4ec5Reranker\u641c\u7d22:\n\u4ec5\u4f7f\u7528Reranker\u641c\u7d22: \u6280\u672f\u6280\u80fd\u548c\u7f16\u7a0b\u8bed\u8a00\n  \u7ed3\u679c1: \u76f8\u5173\u6027=0.0336, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a\u6d4b\u8bd5Bug \u5e76\u8bb0\u5f55\u590d\u73b0\u8fc7\u7a0b\u548c\u6d4b\u8bd5\u7ed3\u679c \uff0c\u5e2e...\n                                                      \u7ed3\u679c2: \u76f8\u5173\u6027=0.0254, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317\u534e\u5927\u5b66-\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u4e13\u4e1a\u3002\u6211\u76ee\u524d\u5df2\u7ecf...\n                                                      \u7ed3\u679c3: \u76f8\u5173\u6027=0.0200, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002-\u3002\u5317\u4eac\u30024. \u8da3\u5934\u6761- \u6d4b\u8bd5\u5b9e\u4e60\u751f\u3002-...                       \n3\ufe0f\u20e3 \u6df7\u5408\u641c\u7d22 (Embedding + Reranker):\n\u6b63\u5728\u8fdb\u884c\u6df7\u5408\u68c0\u7d22: \u6280\u672f\u6280\u80fd\u548c\u7f16\u7a0b\u8bed\u8a00\n==================================================\n\u7b2c\u4e00\u9636\u6bb5\uff1aEmbedding\u7c97\u7b5b\nEmbedding\u9636\u6bb5\u627e\u5230 5 \u4e2a\u5019\u9009\u6587\u6863:\n  \u5019\u90091: \u76f8\u4f3c\u5ea6=0.3679, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317...                                      \u5019\u90092: \u76f8\u4f3c\u5ea6=0.3538, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a...                                \u5019\u90093: \u76f8\u4f3c\u5ea6=0.3130, \u7d22\u5f15=11\n    \u5185\u5bb9: 5. CSDN\/\u963f\u91cc\u4e91\u5f00\u53d1\u8005\u3002\u793e\u533a\u6280\u672f\u535a\u4e3b\u30026. \u963f\u91cc\u4e91\u670d\u52a1\u5668\u8fdb\u9636\u5b9e\u3002\u6218\u8bad\u7ec3\u8425\u30027. \u4e2d\u6d77\u56fd\u9645\u6821\u56ed\u4f01\u4e1a \n\u5b9e\u3002\u8bad\u3002-\u5c0f\u7ec4\u603b\u5206\u7b2c1 \u4f18\u80dc\u30028. \u5927\u5b66\u751f\u201c\u6311\u6218\u676f\u201d\u521b\u3002\u65b0...                                                \u5019\u90094: \u76f8\u4f3c\u5ea6=0.3032, \u7d22\u5f15\n=4\n    \u5185\u5bb9: 11. \u5f00\u53d1\u5de5\u5177\uff1a\u719f\u6089\u5e38\u7528\u7684\u5f00\u53d1\u548c\u7ba1\u7406\u5de5\u5177\u5982\uff1aIDEA\u3001VS\u3001SVN\u3001Github\u3001Jira\u3001Jenkins\u30021. \u57fa\u4e8eWP\nF.UI \u7684Flent \u98ce\u3002\u683c\u684c\u9762\u5e94...                                                                         \u5019\u90095:\n\u76f8\u4f3c\u5ea6=0.2898, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002...                                                        \n\u7b2c\u4e8c\u9636\u6bb5\uff1aReranker\u7cbe\u7b5b\nReranker\u9636\u6bb5\u91cd\u65b0\u6392\u5e8f\u7ed3\u679c:\n  \u6392\u5e8f1: Reranker\u5206\u6570=0.0336, Embedding\u5206\u6570=0.3538, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a...                                \u6392\u5e8f2: Reranker\u5206\u6570=0.0254, Embed\nding\u5206\u6570=0.3679, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317...                                      \u6392\u5e8f3: Reranker\u5206\u6570=0.0200, Em\nbedding\u5206\u6570=0.2898, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002...                                                          \u6392\u5e8f4: Reranker\u5206\u6570\n=0.0115, Embedding\u5206\u6570=0.3130, \u7d22\u5f15=11\n    \u5185\u5bb9: 5. CSDN\/\u963f\u91cc\u4e91\u5f00\u53d1\u8005\u3002\u793e\u533a\u6280\u672f\u535a\u4e3b\u30026. \u963f\u91cc\u4e91\u670d\u52a1\u5668\u8fdb\u9636\u5b9e\u3002\u6218\u8bad\u7ec3\u8425\u30027. \u4e2d\u6d77\u56fd\u9645\u6821\u56ed\u4f01\u4e1a \n\u5b9e\u3002\u8bad\u3002-\u5c0f\u7ec4\u603b\u5206\u7b2c1 \u4f18\u80dc\u30028. \u5927\u5b66\u751f\u201c\u6311\u6218\u676f\u201d\u521b\u3002\u65b0...                                                \u6392\u5e8f5: Reranker\u5206\u6570=0.0\n003, Embedding\u5206\u6570=0.3032, \u7d22\u5f15=4\n    \u5185\u5bb9: 11. \u5f00\u53d1\u5de5\u5177\uff1a\u719f\u6089\u5e38\u7528\u7684\u5f00\u53d1\u548c\u7ba1\u7406\u5de5\u5177\u5982\uff1aIDEA\u3001VS\u3001SVN\u3001Github\u3001Jira\u3001Jenkins\u30021. \u57fa\u4e8eWP\nF.UI \u7684Flent \u98ce\u3002\u683c\u684c\u9762\u5e94...                                                                         \u7ed3\u679c1:\n\u6700\u7ec8\u76f8\u5173\u6027=0.0336, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a\u6d4b\u8bd5Bug \u5e76\u8bb0\u5f55\u590d\u73b0\u8fc7\u7a0b\u548c\u6d4b\u8bd5\u7ed3\u679c \uff0c\u5e2e...\n                                                      \u7ed3\u679c2: \u6700\u7ec8\u76f8\u5173\u6027=0.0254, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317\u534e\u5927\u5b66-\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u4e13\u4e1a\u3002\u6211\u76ee\u524d\u5df2\u7ecf...\n                                                      \u7ed3\u679c3: \u6700\u7ec8\u76f8\u5173\u6027=0.0200, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002-\u3002\u5317\u4eac\u30024. \u8da3\u5934\u6761- \u6d4b\u8bd5\u5b9e\u4e60\u751f\u3002-...                       \n============================================================\n\u67e5\u8be2: \u5e94\u8058\u7684\u5c97\u4f4d\u548c\u516c\u53f8\n============================================================\n\n1\ufe0f\u20e3 \u4ec5Embedding\u641c\u7d22:\n\u4ec5\u4f7f\u7528Embedding\u641c\u7d22: \u5e94\u8058\u7684\u5c97\u4f4d\u548c\u516c\u53f8\n  \u7ed3\u679c1: \u76f8\u4f3c\u5ea6=0.4570, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002-\u3002\u5317\u4eac\u30024. \u8da3\u5934\u6761- \u6d4b\u8bd5\u5b9e\u4e60\u751f\u3002-...                         \u7ed3\u679c2: \u76f8\u4f3c\u5ea6=0.4546, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317\u534e\u5927\u5b66-\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u4e13\u4e1a\u3002\u6211\u76ee\u524d\u5df2\u7ecf...\n                                                      \u7ed3\u679c3: \u76f8\u4f3c\u5ea6=0.3887, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a\u6d4b\u8bd5Bug \u5e76\u8bb0\u5f55\u590d\u73b0\u8fc7\u7a0b\u548c\u6d4b\u8bd5\u7ed3\u679c \uff0c\u5e2e...\n                                                    \n2\ufe0f\u20e3 \u4ec5Reranker\u641c\u7d22:\n\u4ec5\u4f7f\u7528Reranker\u641c\u7d22: \u5e94\u8058\u7684\u5c97\u4f4d\u548c\u516c\u53f8\n  \u7ed3\u679c1: \u76f8\u5173\u6027=0.9932, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317\u534e\u5927\u5b66-\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u4e13\u4e1a\u3002\u6211\u76ee\u524d\u5df2\u7ecf...\n                                                      \u7ed3\u679c2: \u76f8\u5173\u6027=0.4961, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. Deep - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002-\u3002\u5317\u4eac\u30024. \u8da3\u5934\u6761- \u6d4b\u8bd5\u5b9e\u4e60\u751f\u3002-...                         \u7ed3\u679c3: \u76f8\u5173\u6027=0.0013, \u7d22\u5f15=11\n    \u5185\u5bb9: 5. CSDN\/\u963f\u91cc\u4e91\u5f00\u53d1\u8005\u3002\u793e\u533a\u6280\u672f\u535a\u4e3b\u30026. \u963f\u91cc\u4e91\u670d\u52a1\u5668\u8fdb\u9636\u5b9e\u3002\u6218\u8bad\u7ec3\u8425\u30027. \u4e2d\u6d77\u56fd\u9645\u6821\u56ed\u4f01\u4e1a \n\u5b9e\u3002\u8bad\u3002-\u5c0f\u7ec4\u603b\u5206\u7b2c1 \u4f18\u80dc\u30028. \u5927\u5b66\u751f\u201c\u6311\u6218\u676f\u201d\u521b\u3002\u65b0\u521b\u4e1a\u5927\u8d5b-\u6821\u7ea7\u4e8c\u7b49\u3002\u5956\u3001\u7701\u7ea7\u4e09\u7b49\u5956\u30029....         \n3\ufe0f\u20e3 \u6df7\u5408\u641c\u7d22 (Embedding + Reranker):\n\u6b63\u5728\u8fdb\u884c\u6df7\u5408\u68c0\u7d22: \u5e94\u8058\u7684\u5c97\u4f4d\u548c\u516c\u53f8\n==================================================\n\u7b2c\u4e00\u9636\u6bb5\uff1aEmbedding\u7c97\u7b5b\nEmbedding\u9636\u6bb5\u627e\u5230 5 \u4e2a\u5019\u9009\u6587\u6863:\n  \u5019\u90091: \u76f8\u4f3c\u5ea6=0.4570, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. Deep - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002...                                                          \u5019\u90092: \u76f8\u4f3c\u5ea6=0.4546\n, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317...                                      \u5019\u90093: \u76f8\u4f3c\u5ea6=0.3887, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a...                                \u5019\u90094: \u76f8\u4f3c\u5ea6=0.3584, \u7d22\u5f15=1\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u56fe\u50cf\u5904\u7406\u4e0e\u6807\u6ce8\u8f6f\u4ef6\u5957\u88c5\u7684\u5f00\u53d1\u4e0e\u6d4b\u8bd5\u3002\u5bf9\u63a5\u5ba2\u6237\u9700\u6c42\uff0c\u8d1f\u8d23\u8f6f\u4ef6\u67b6\u6784\u8bbe\u3002\u8ba1\uff0c\u8bbe\u8ba1UI \u539f\u578b\uff0c \n\u8c03\u7814\u76f8\u540c\u7ade\u54c1\u5f62\u6210\u62a5\u544a\uff0c\u786e\u5b9a\u8f6f\u4ef6\u6280\u672f\u67b6\u6784\u548c\u6a21\u5757\u529f\u80fd\u5212\u5206\u3002\u4e3b\u3002\u5bfc\u63a7\u5236\u6a21\u5757...                              \u5019\u90095: \u76f8\u4f3c\u5ea6=0.3318, \u7d22\u5f15=3\n    \u5185\u5bb9: 4. AI \u5927\u6a21\u578b\uff1a\u638c\u63e1\u7b49\u5927\u6a21\u578b\u90e8\u7f72\u4e0e\u8c03\u4f18\uff0c\u719f\u6089\u534f\u8bae\u5b9a\u4e49\u4e0e\u63a5\u53e3\u5f00\u53d1\u4f7f\u7528\u30025. \u6d4b\u8bd5\u6280\u80fd\uff1a \n\u719f\u6089\u5355\u5143\u6d4b\u8bd5\u3001\u63a5\u53e3\u6d4b\u8bd5\u3001UI \u6d4b\u8bd5\u7684\u6279\u5904\u7406\u4e0e\u81ea\u52a8\u5316\u811a\u672c...                                             \n\u7b2c\u4e8c\u9636\u6bb5\uff1aReranker\u7cbe\u7b5b\nReranker\u9636\u6bb5\u91cd\u65b0\u6392\u5e8f\u7ed3\u679c:\n  \u6392\u5e8f1: Reranker\u5206\u6570=0.9932, Embedding\u5206\u6570=0.4546, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317...                                      \u6392\u5e8f2: Reranker\u5206\u6570=0.5000, Em\nbedding\u5206\u6570=0.4570, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002...                                                          \u6392\u5e8f3: Reranker\u5206\u6570\n=0.0011, Embedding\u5206\u6570=0.3887, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a...                                \u6392\u5e8f4: Reranker\u5206\u6570=0.0002, Embed\nding\u5206\u6570=0.3584, \u7d22\u5f15=1\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u56fe\u50cf\u5904\u7406\u4e0e\u6807\u6ce8\u8f6f\u4ef6\u5957\u88c5\u7684\u5f00\u53d1\u4e0e\u6d4b\u8bd5\u3002\u5bf9\u63a5\u5ba2\u6237\u9700\u6c42\uff0c\u8d1f\u8d23\u8f6f\u4ef6\u67b6\u6784\u8bbe\u3002\u8ba1\uff0c\u8bbe\u8ba1UI \u539f\u578b\uff0c \n\u8c03\u7814\u76f8\u540c\u7ade\u54c1\u5f62\u6210\u62a5\u544a\uff0c\u786e\u5b9a\u8f6f\u4ef6\u6280\u672f\u67b6\u6784\u548c\u6a21\u5757\u529f\u80fd\u5212\u5206\u3002\u4e3b\u3002\u5bfc\u63a7\u5236\u6a21\u5757...                              \u6392\u5e8f5: Reranker\u5206\u6570=0.0001, Embeddin\ng\u5206\u6570=0.3318, \u7d22\u5f15=3\n    \u5185\u5bb9: 4. AI \u5927\u6a21\u578b\uff1a\u638c\u63e1DeepSeek \u7b49\u5927\u6a21\u578b\u90e8\u7f72\u4e0e\u8c03\u4f18\uff0c\u719f\u6089\u534f\u8bae\u5b9a\u4e49\u4e0e\u63a5\u53e3\u5f00\u53d1\u4f7f\u7528\u30025. \u6d4b\u8bd5\u6280\u80fd\uff1a \n\u719f\u6089\u5355\u5143\u6d4b\u8bd5\u3001\u63a5\u53e3\u6d4b\u8bd5\u3001UI \u6d4b\u8bd5\u7684\u6279\u5904\u7406\u4e0e\u81ea\u52a8\u5316\u811a\u672c...                                               \u7ed3\u679c1: \u6700\u7ec8\u76f8\u5173\u6027=0.9932, \u7d22\u5f15=\n0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317**\u5b66-\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u4e13\u4e1a\u3002\u6211\u76ee\u524d\u5df2\u7ecf...\n                                                      \u7ed3\u679c2: \u6700\u7ec8\u76f8\u5173\u6027=0.5000, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepA*** - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002-\u3002\u5317\u4eac\u30024. \u8da3\u5934\u6761- \u6d4b\u8bd5\u5b9e\u4e60\u751f\u3002-...                         \u7ed3\u679c3: \u6700\u7ec8\u76f8\u5173\u6027=0.0011, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a\u6d4b\u8bd5Bug \u5e76\u8bb0\u5f55\u590d\u73b0\u8fc7\u7a0b\u548c\u6d4b\u8bd5\u7ed3\u679c \uff0c\u5e2e...\n                                                    \n============================================================\n\u67e5\u8be2: \u6d4b\u8bd5\u548c\u5f00\u53d1\u7ecf\u9a8c\n============================================================\n\n1\ufe0f\u20e3 \u4ec5Embedding\u641c\u7d22:\n\u4ec5\u4f7f\u7528Embedding\u641c\u7d22: \u6d4b\u8bd5\u548c\u5f00\u53d1\u7ecf\u9a8c\n  \u7ed3\u679c1: \u76f8\u4f3c\u5ea6=0.5181, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a\u6d4b\u8bd5Bug \u5e76\u8bb0\u5f55\u590d\u73b0\u8fc7\u7a0b\u548c\u6d4b\u8bd5\u7ed3\u679c \uff0c\u5e2e...\n                                                      \u7ed3\u679c2: \u76f8\u4f3c\u5ea6=0.5156, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002-\u3002\u5317\u4eac\u30024. \u8da3\u5934\u6761- \u6d4b\u8bd5\u5b9e\u4e60\u751f\u3002-...                         \u7ed3\u679c3: \u76f8\u4f3c\u5ea6=0.4897, \u7d22\u5f15=3\n    \u5185\u5bb9: 4. AI \u5927\u6a21\u578b\uff1a\u638c\u63e1DeepSeek \u7b49\u5927\u6a21\u578b\u90e8\u7f72\u4e0e\u8c03\u4f18\uff0c\u719f\u6089\u534f\u8bae\u5b9a\u4e49\u4e0e\u63a5\u53e3\u5f00\u53d1\u4f7f\u7528\u30025. \u6d4b\u8bd5\u6280\u80fd\uff1a \n\u719f\u6089\u5355\u5143\u6d4b\u8bd5\u3001\u63a5\u53e3\u6d4b\u8bd5\u3001UI \u6d4b\u8bd5\u7684\u6279\u5904\u7406\u4e0e\u81ea\u52a8\u5316\u811a\u672c\u30026. \u6d4b\u8bd5\u5de5\u5177\uff1a\u719f\u6089\u6d4b\u8bd5\u6d41\u7a0b\u3001\u6d4b\u8bd5\u6846\u67b6...        \n2\ufe0f\u20e3 \u4ec5Reranker\u641c\u7d22:\n\u4ec5\u4f7f\u7528Reranker\u641c\u7d22: \u6d4b\u8bd5\u548c\u5f00\u53d1\u7ecf\u9a8c\n  \u7ed3\u679c1: \u76f8\u5173\u6027=0.9741, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002-\u3002\u5317\u4eac\u30024. \u8da3\u5934\u6761- \u6d4b\u8bd5\u5b9e\u4e60\u751f\u3002-...                         \u7ed3\u679c2: \u76f8\u5173\u6027=0.9541, \u7d22\u5f15=3\n    \u5185\u5bb9: 4. AI \u5927\u6a21\u578b\uff1a\u638c\u63e1DeepSeek \u7b49\u5927\u6a21\u578b\u90e8\u7f72\u4e0e\u8c03\u4f18\uff0c\u719f\u6089\u534f\u8bae\u5b9a\u4e49\u4e0e\u63a5\u53e3\u5f00\u53d1\u4f7f\u7528\u30025. \u6d4b\u8bd5\u6280\u80fd\uff1a \n\u719f\u6089\u5355\u5143\u6d4b\u8bd5\u3001\u63a5\u53e3\u6d4b\u8bd5\u3001UI \u6d4b\u8bd5\u7684\u6279\u5904\u7406\u4e0e\u81ea\u52a8\u5316\u811a\u672c\u30026. \u6d4b\u8bd5\u5de5\u5177\uff1a\u719f\u6089\u6d4b\u8bd5\u6d41\u7a0b\u3001\u6d4b\u8bd5\u6846\u67b6...          \u7ed3\u679c3: \u76f8\u5173\u6027=0.8872, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a\u6d4b\u8bd5Bug \u5e76\u8bb0\u5f55\u590d\u73b0\u8fc7\u7a0b\u548c\u6d4b\u8bd5\u7ed3\u679c \uff0c\u5e2e...\n                                                    \n3\ufe0f\u20e3 \u6df7\u5408\u641c\u7d22 (Embedding + Reranker):\n\u6b63\u5728\u8fdb\u884c\u6df7\u5408\u68c0\u7d22: \u6d4b\u8bd5\u548c\u5f00\u53d1\u7ecf\u9a8c\n==================================================\n\u7b2c\u4e00\u9636\u6bb5\uff1aEmbedding\u7c97\u7b5b\nEmbedding\u9636\u6bb5\u627e\u5230 5 \u4e2a\u5019\u9009\u6587\u6863:\n  \u5019\u90091: \u76f8\u4f3c\u5ea6=0.5181, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a...                                \u5019\u90092: \u76f8\u4f3c\u5ea6=0.5156, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002...                                                          \u5019\u90093: \u76f8\u4f3c\u5ea6=0.4897\n, \u7d22\u5f15=3\n    \u5185\u5bb9: 4. AI \u5927\u6a21\u578b\uff1a\u638c\u63e1DeepSeek \u7b49\u5927\u6a21\u578b\u90e8\u7f72\u4e0e\u8c03\u4f18\uff0c\u719f\u6089\u534f\u8bae\u5b9a\u4e49\u4e0e\u63a5\u53e3\u5f00\u53d1\u4f7f\u7528\u30025. \u6d4b\u8bd5\u6280\u80fd\uff1a \n\u719f\u6089\u5355\u5143\u6d4b\u8bd5\u3001\u63a5\u53e3\u6d4b\u8bd5\u3001UI \u6d4b\u8bd5\u7684\u6279\u5904\u7406\u4e0e\u81ea\u52a8\u5316\u811a\u672c...                                               \u5019\u90094: \u76f8\u4f3c\u5ea6=0.4097, \u7d22\u5f15=1\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u56fe\u50cf\u5904\u7406\u4e0e\u6807\u6ce8\u8f6f\u4ef6\u5957\u88c5\u7684\u5f00\u53d1\u4e0e\u6d4b\u8bd5\u3002\u5bf9\u63a5\u5ba2\u6237\u9700\u6c42\uff0c\u8d1f\u8d23\u8f6f\u4ef6\u67b6\u6784\u8bbe\u3002\u8ba1\uff0c\u8bbe\u8ba1UI \u539f\u578b\uff0c \n\u8c03\u7814\u76f8\u540c\u7ade\u54c1\u5f62\u6210\u62a5\u544a\uff0c\u786e\u5b9a\u8f6f\u4ef6\u6280\u672f\u67b6\u6784\u548c\u6a21\u5757\u529f\u80fd\u5212\u5206\u3002\u4e3b\u3002\u5bfc\u63a7\u5236\u6a21\u5757...                              \u5019\u90095: \u76f8\u4f3c\u5ea6=0.3918, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317...                                    \n\u7b2c\u4e8c\u9636\u6bb5\uff1aReranker\u7cbe\u7b5b\nReranker\u9636\u6bb5\u91cd\u65b0\u6392\u5e8f\u7ed3\u679c:\n  \u6392\u5e8f1: Reranker\u5206\u6570=0.9736, Embedding\u5206\u6570=0.5156, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002...                                                          \u6392\u5e8f2: Reranker\u5206\u6570\n=0.9541, Embedding\u5206\u6570=0.4897, \u7d22\u5f15=3\n    \u5185\u5bb9: 4. AI \u5927\u6a21\u578b\uff1a\u638c\u63e1DeepSeek \u7b49\u5927\u6a21\u578b\u90e8\u7f72\u4e0e\u8c03\u4f18\uff0c\u719f\u6089\u534f\u8bae\u5b9a\u4e49\u4e0e\u63a5\u53e3\u5f00\u53d1\u4f7f\u7528\u30025. \u6d4b\u8bd5\u6280\u80fd\uff1a \n\u719f\u6089\u5355\u5143\u6d4b\u8bd5\u3001\u63a5\u53e3\u6d4b\u8bd5\u3001UI \u6d4b\u8bd5\u7684\u6279\u5904\u7406\u4e0e\u81ea\u52a8\u5316\u811a\u672c...                                               \u6392\u5e8f3: Reranker\u5206\u6570=0.8882\n, Embedding\u5206\u6570=0.5181, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a...                                \u6392\u5e8f4: Reranker\u5206\u6570=0.8223, Embed\nding\u5206\u6570=0.4097, \u7d22\u5f15=1\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u56fe\u50cf\u5904\u7406\u4e0e\u6807\u6ce8\u8f6f\u4ef6\u5957\u88c5\u7684\u5f00\u53d1\u4e0e\u6d4b\u8bd5\u3002\u5bf9\u63a5\u5ba2\u6237\u9700\u6c42\uff0c\u8d1f\u8d23\u8f6f\u4ef6\u67b6\u6784\u8bbe\u3002\u8ba1\uff0c\u8bbe\u8ba1UI \u539f\u578b\uff0c \n\u8c03\u7814\u76f8\u540c\u7ade\u54c1\u5f62\u6210\u62a5\u544a\uff0c\u786e\u5b9a\u8f6f\u4ef6\u6280\u672f\u67b6\u6784\u548c\u6a21\u5757\u529f\u80fd\u5212\u5206\u3002\u4e3b\u3002\u5bfc\u63a7\u5236\u6a21\u5757...                              \u6392\u5e8f5: Reranker\u5206\u6570=0.0014, Embeddin\ng\u5206\u6570=0.3918, \u7d22\u5f15=0\n    \u5185\u5bb9: \u7b2c1\u9875: \u6280\u672f\u6808\u3002Boss \u60a8\u597d\u3002\u6211\u53eb\u674e\u660e\uff0c\u6709\u610f\u5e94\u8058\u8d35\u516c\u53f8\uff1a\u540e\u7aef\u5f00\u53d1\u3001\u5168\u6808\u5f00\u53d1\u5de5\u7a0b\u5e08\u3001\u4ea7\u54c1\u7ecf\u7406 \n\u7b49\u5c97\u4f4d\u3002\u6211\u62e5\u6709\u5b9e\u9645\u5de5\u4f5c\u5f00\u53d1\u7ecf\u9a8c3-5 \u5e74\uff0c\u64c5\u957f\u5168\u6808\u5f00\u53d1\u3002\u6bd5\u4e1a\u4e8e\u5317...                                      \u7ed3\u679c1: \u6700\u7ec8\u76f8\u5173\u6027=0.9736, \u7d22\u5f15=10\n    \u5185\u5bb9: ELK \u65e5\u5fd7\u7cfb\u7edf\u5b9e\u65f6\u91c7\u96c6\u5f02\u5e38\u4fe1\u606f\u30021. DeepAgile - \u8f6f\u4ef6\u5f00\u53d1\u5de5\u7a0b\u5e08\u3002-\u3002\u82cf\u5dde\u30022. \u6ef4\u6ef4- \u6d4b\u8bd5\u5f00\u53d1\u5b9e\n\u4e60\u751f\u3002-\u3002\u676d\u5dde\u30023. \u7528\u53cb- \u8fd0\u7ef4\u5f00\u53d1\u5b9e\u4e60\u751f\u3002-\u3002\u5317\u4eac\u30024. \u8da3\u5934\u6761- \u6d4b\u8bd5\u5b9e\u4e60\u751f\u3002-...                         \u7ed3\u679c2: \u6700\u7ec8\u76f8\u5173\u6027=0.9541, \u7d22\u5f15=3\n    \u5185\u5bb9: 4. AI \u5927\u6a21\u578b\uff1a\u638c\u63e1DeepSeek \u7b49\u5927\u6a21\u578b\u90e8\u7f72\u4e0e\u8c03\u4f18\uff0c\u719f\u6089\u534f\u8bae\u5b9a\u4e49\u4e0e\u63a5\u53e3\u5f00\u53d1\u4f7f\u7528\u30025. \u6d4b\u8bd5\u6280\u80fd\uff1a \n\u719f\u6089\u5355\u5143\u6d4b\u8bd5\u3001\u63a5\u53e3\u6d4b\u8bd5\u3001UI \u6d4b\u8bd5\u7684\u6279\u5904\u7406\u4e0e\u81ea\u52a8\u5316\u811a\u672c\u30026. \u6d4b\u8bd5\u5de5\u5177\uff1a\u719f\u6089\u6d4b\u8bd5\u6d41\u7a0b\u3001\u6d4b\u8bd5\u6846\u67b6...          \u7ed3\u679c3: \u6700\u7ec8\u76f8\u5173\u6027=0.8882, \u7d22\u5f15=2\n    \u5185\u5bb9: \u00b7\u53c2\u4e0e\u4ea7\u54c1\u7814\u53d1\u8fc7\u7a0b\u8d28\u91cf\u63a7\u5236\u4e0e\u7ba1\u7406\uff0c\u8d1f\u8d23\u767d\u76d2\u6d4b\u8bd5\u4e0e\u6027\u80fd\u6d4b\u8bd5\u4e0e\u5206\u6790\u3002\u7ef4\u62a4\u548c\u3002\u7f16\u5199\u63a5\u53e3\u6587\u6863\u3002\u6839\u636e\n\u6d4b\u8bd5\u8ba1\u5212\u548c\u8981\u6c42\u7f16\u5199\u6d4b\u8bd5\u7528\u4f8b\u5bf9Web \u7aef\u8fdb\u884c\u81ea\u52a8\u5316\u6d4b\u8bd5\u548c\u63a5\u53e3\u6d4b\u3002\u8bd5\u3002\u8ddf\u8e2a\u6d4b\u8bd5Bug \u5e76\u8bb0\u5f55\u590d\u73b0\u8fc7\u7a0b\u548c\u6d4b\u8bd5\u7ed3\u679c \uff0c\u5e2e...\n                                                    <\/code><\/pre>\n\n\n\n
      \ud83c\udfaf \u6df7\u5408\u68c0\u7d22\u7cfb\u7edf\u6210\u529f\u8fd0\u884c\uff01<\/h2>\n\n\n\n
      \u5206\u6790Embedding vs Reranker<\/strong>\u7684\u533a\u522b\u548c\u6548\u679c\uff1a<\/p>\n\n\n\n
      \ud83d\udcca \u6a21\u578b\u5bf9\u6bd4\u5206\u6790<\/h2>\n\n\n\n
      1. Embedding\u6a21\u578b\u7279\u70b9<\/strong><\/h3>\n\n\n\n
        \n
      • \u901f\u5ea6<\/strong>: \u26a1 \u5feb\u901f\uff08\u5411\u91cf\u76f8\u4f3c\u5ea6\u8ba1\u7b97\uff09<\/li>\n\n\n\n
      • \u7cbe\u5ea6<\/strong>: \ufffd\ufffd \u4e2d\u7b49\uff08\u8bed\u4e49\u76f8\u4f3c\u5ea6\uff09<\/li>\n\n\n\n
      • \u9002\u7528\u573a\u666f<\/strong>: \u7c97\u7b5b\u9636\u6bb5\uff0c\u4ece\u5927\u91cf\u6587\u6863\u4e2d\u5feb\u901f\u7b5b\u9009\u5019\u9009<\/li>\n<\/ul>\n\n\n\n
        2. Reranker\u6a21\u578b\u7279\u70b9<\/strong><\/h3>\n\n\n\n
          \n
        • \u901f\u5ea6<\/strong>: \ud83d\udc0c \u8f83\u6162\uff08\u9700\u8981\u9010\u5bf9\u8ba1\u7b97\u76f8\u5173\u6027\uff09<\/li>\n\n\n\n
        • \u7cbe\u5ea6<\/strong>: \u9ad8\u7cbe\u5ea6\uff08\u4e13\u95e8\u7684\u76f8\u5173\u6027\u5224\u65ad\uff09<\/li>\n\n\n\n
        • \u9002\u7528\u573a\u666f<\/strong>: \u7cbe\u7b5b\u9636\u6bb5\uff0c\u5bf9\u5019\u9009\u6587\u6863\u8fdb\u884c\u7cbe\u786e\u6392\u5e8f<\/li>\n<\/ul>\n\n\n\n
          \ud83d\udd0d \u5b9e\u9645\u6548\u679c\u5bf9\u6bd4<\/h2>\n\n\n\n
          \u67e5\u8be21: “\u674e\u660e\u7684\u5de5\u4f5c\u7ecf\u9a8c”<\/strong><\/h3>\n\n\n\n
          \u65b9\u6cd5<\/th>\u7ed3\u679c1<\/th>\u7ed3\u679c2<\/th>\u7ed3\u679c3<\/th><\/tr><\/thead>
          \u4ec5Embedding<\/strong><\/td>\u76f8\u4f3c\u5ea6=0.4475 (\u81ea\u6211\u4ecb\u7ecd)<\/td>\u76f8\u4f3c\u5ea6=0.3059 (\u6d4b\u8bd5\u7ecf\u9a8c)<\/td>\u76f8\u4f3c\u5ea6=0.2979 (\u5f00\u53d1\u7ecf\u9a8c)<\/td><\/tr>
          \u4ec5Reranker<\/strong><\/td>\u76f8\u5173\u6027=0.9995 (\u81ea\u6211\u4ecb\u7ecd)<\/td>\u76f8\u5173\u6027=0.9819 (\u5de5\u4f5c\u7ecf\u5386)<\/td>\u76f8\u5173\u6027=0.0158 (\u793e\u4f1a\u5b9e\u8df5)<\/td><\/tr>
          \u6df7\u5408\u68c0\u7d22<\/strong><\/td>\u76f8\u5173\u6027=0.9995 (\u81ea\u6211\u4ecb\u7ecd)<\/td>\u76f8\u5173\u6027=0.9819 (\u5de5\u4f5c\u7ecf\u5386)<\/td>\u76f8\u5173\u6027=0.0028 (\u5f00\u53d1\u7ecf\u9a8c)<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
          \u67e5\u8be22: “\u6280\u672f\u6280\u80fd\u548c\u7f16\u7a0b\u8bed\u8a00”<\/strong><\/h3>\n\n\n\n
          \u65b9\u6cd5<\/th>\u7ed3\u679c1<\/th>\u7ed3\u679c2<\/th>\u7ed3\u679c3<\/th><\/tr><\/thead>
          \u4ec5Embedding<\/strong><\/td>\u76f8\u4f3c\u5ea6=0.3679 (\u81ea\u6211\u4ecb\u7ecd)<\/td>\u76f8\u4f3c\u5ea6=0.3538 (\u6d4b\u8bd5\u7ecf\u9a8c)<\/td>\u76f8\u4f3c\u5ea6=0.3130 (\u793e\u4f1a\u5b9e\u8df5)<\/td><\/tr>
          \u4ec5Reranker<\/strong><\/td>\u76f8\u5173\u6027=0.0336 (\u6d4b\u8bd5\u7ecf\u9a8c)<\/td>\u76f8\u5173\u6027=0.0254 (\u81ea\u6211\u4ecb\u7ecd)<\/td>\u76f8\u5173\u6027=0.0200 (\u5de5\u4f5c\u7ecf\u5386)<\/td><\/tr>
          \u6df7\u5408\u68c0\u7d22<\/strong><\/td>\u76f8\u5173\u6027=0.0336 (\u6d4b\u8bd5\u7ecf\u9a8c)<\/td>\u76f8\u5173\u6027=0.0254 (\u81ea\u6211\u4ecb\u7ecd)<\/td>\u76f8\u5173\u6027=0.0200 (\u5de5\u4f5c\u7ecf\u5386)<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
          \ud83d\ude80 \u5173\u952e\u53d1\u73b0<\/h2>\n\n\n\n
          1. Reranker\u7684\u7cbe\u51c6\u6027<\/strong><\/h3>\n\n\n\n
            \n
          • \u9ad8\u76f8\u5173\u6027\u5224\u65ad<\/strong>: \u5bf9\u4e8e\u660e\u786e\u76f8\u5173\u7684\u67e5\u8be2\uff0cReranker\u80fd\u7ed9\u51fa\u63a5\u8fd11.0\u7684\u9ad8\u5206<\/li>\n\n\n\n
          • \u4f4e\u76f8\u5173\u6027\u8fc7\u6ee4<\/strong>: \u5bf9\u4e8e\u4e0d\u76f8\u5173\u7684\u5185\u5bb9\uff0cReranker\u80fd\u7ed9\u51fa\u63a5\u8fd10.0\u7684\u4f4e\u5206<\/li>\n\n\n\n
          • \u533a\u5206\u5ea6\u5f3a<\/strong>: \u5206\u6570\u5206\u5e03\u66f4\u52a0\u6781\u7aef\uff0c\u4fbf\u4e8e\u533a\u5206\u76f8\u5173\u548c\u4e0d\u76f8\u5173\u5185\u5bb9<\/li>\n<\/ul>\n\n\n\n
            2. Embedding\u7684\u8bed\u4e49\u7406\u89e3<\/strong><\/h3>\n\n\n\n
              \n
            • \u8bed\u4e49\u76f8\u4f3c<\/strong>: \u80fd\u627e\u5230\u8bed\u4e49\u76f8\u5173\u4f46\u4e0d\u5b8c\u5168\u5339\u914d\u7684\u5185\u5bb9<\/li>\n\n\n\n
            • \u5206\u6570\u5e73\u6ed1<\/strong>: \u76f8\u4f3c\u5ea6\u5206\u6570\u5206\u5e03\u76f8\u5bf9\u5e73\u6ed1<\/li>\n\n\n\n
            • \u53ec\u56de\u7387\u9ad8<\/strong>: \u80fd\u627e\u5230\u66f4\u591a\u6f5c\u5728\u76f8\u5173\u7684\u5185\u5bb9<\/li>\n<\/ul>\n\n\n\n
              3. \u6df7\u5408\u68c0\u7d22\u7684\u4f18\u52bf<\/strong><\/h3>\n\n\n\n
                \n
              • \u6548\u7387\u5e73\u8861<\/strong>: Embedding\u5feb\u901f\u7b5b\u9009\u5019\u9009\uff0cReranker\u7cbe\u786e\u6392\u5e8f<\/li>\n\n\n\n
              • \u7cbe\u5ea6\u63d0\u5347<\/strong>: \u6700\u7ec8\u7ed3\u679c\u8d28\u91cf\u663e\u8457\u9ad8\u4e8e\u5355\u72ec\u4f7f\u7528\u4efb\u4e00\u6a21\u578b<\/li>\n\n\n\n
              • \u53ef\u6269\u5c55\u6027<\/strong>: \u9002\u5408\u5927\u89c4\u6a21\u6587\u6863\u68c0\u7d22\u573a\u666f<\/li>\n<\/ul>\n\n\n\n
                \ud83d\udca1 \u4f7f\u7528\u5efa\u8bae<\/h2>\n\n\n\n
                \u4f55\u65f6\u4f7f\u7528Embedding<\/strong><\/h3>\n\n\n\n
                  \n
                • \u6587\u6863\u5e93\u5f88\u5927\uff08>1000\u7bc7\uff09<\/li>\n\n\n\n
                • \u9700\u8981\u5feb\u901f\u54cd\u5e94<\/li>\n\n\n\n
                • \u5bf9\u7cbe\u5ea6\u8981\u6c42\u4e0d\u662f\u7279\u522b\u9ad8<\/li>\n<\/ul>\n\n\n\n
                  \u4f55\u65f6\u4f7f\u7528Reranker<\/strong><\/h3>\n\n\n\n
                    \n
                  • \u6587\u6863\u5e93\u8f83\u5c0f\uff08<1000\u7bc7\uff09<\/li>\n\n\n\n
                  • \u5bf9\u7cbe\u5ea6\u8981\u6c42\u5f88\u9ad8<\/li>\n\n\n\n
                  • \u53ef\u4ee5\u63a5\u53d7\u8f83\u6162\u7684\u54cd\u5e94\u901f\u5ea6<\/li>\n<\/ul>\n\n\n\n
                    \u4f55\u65f6\u4f7f\u7528\u6df7\u5408\u68c0\u7d22<\/strong><\/h3>\n\n\n\n
                      \n
                    • \u751f\u4ea7\u73af\u5883\u7684\u6700\u4f73\u9009\u62e9<\/li>\n\n\n\n
                    • \u65e2\u9700\u8981\u901f\u5ea6\u53c8\u9700\u8981\u7cbe\u5ea6<\/li>\n\n\n\n
                    • \u6587\u6863\u5e93\u89c4\u6a21\u4e2d\u7b49<\/li>\n<\/ul>\n\n\n\n
                      \u6df7\u5408\u68c0\u7d22\u7cfb\u7edf\u5df2\u7ecf\u5b8c\u7f8e\u5c55\u793a\u4e86\u4e24\u79cd\u6a21\u578b\u7684\u4f18\u52bf\u548c\u4e92\u8865\u6027\uff0c\u5728\u5b9e\u9645\u5e94\u7528\u4e2d\u975e\u5e38\u6709\u6548\u3002<\/p>\n\n\n\n
                      import fitz  # PyMuPDF\nimport numpy as np\nfrom typing import List, Tuple, Dict\nfrom test_qwen3_embedding import Qwen3Embedding\nfrom test_qwen3_reranker import Qwen3Reranker\nimport torch\nimport re\n\nclass HybridPDFRetriever:\n    def __init__(self, \n                 embedding_model_path: str = \"models\/Qwen3-Embedding-0.6B\/Qwen\/Qwen3-Embedding-0.6B\",\n                 reranker_model_path: str = \"models\/Qwen3-Reranker-0.6B\/Qwen\/Qwen3-Reranker-0.6B\"):\n        \"\"\"\n        \u521d\u59cb\u5316\u6df7\u5408PDF\u68c0\u7d22\u5668\n        Args:\n            embedding_model_path: Qwen3 Embedding\u6a21\u578b\u8def\u5f84\n            reranker_model_path: Qwen3 Reranker\u6a21\u578b\u8def\u5f84\n        \"\"\"\n        print(\"\u6b63\u5728\u52a0\u8f7dQwen3 Embedding\u6a21\u578b...\")\n        self.embedding_model = Qwen3Embedding(embedding_model_path)\n        print(\"Embedding\u6a21\u578b\u52a0\u8f7d\u5b8c\u6210\uff01\")\n        \n        print(\"\u6b63\u5728\u52a0\u8f7dQwen3 Reranker\u6a21\u578b...\")\n        self.reranker_model = Qwen3Reranker(\n            model_name_or_path=reranker_model_path,\n            instruction=\"Given the user query, retrieval the relevant passages\",\n            max_length=2048\n        )\n        print(\"Reranker\u6a21\u578b\u52a0\u8f7d\u5b8c\u6210\uff01\")\n        \n        self.documents = []\n        self.embeddings = None\n        self.chunk_size = 300\n        \n    def load_pdf(self, pdf_path: str) -> List[str]:\n        \"\"\"\u52a0\u8f7dPDF\u6587\u4ef6\u5e76\u63d0\u53d6\u6587\u672c\"\"\"\n        print(f\"\u6b63\u5728\u52a0\u8f7dPDF\u6587\u4ef6: {pdf_path}\")\n        \n        try:\n            doc = fitz.open(pdf_path)\n            full_text = \"\"\n            \n            for page_num in range(len(doc)):\n                page = doc.load_page(page_num)\n                text = page.get_text()\n                full_text += f\"\u7b2c{page_num + 1}\u9875: \" + text + \"\\n\"\n            \n            doc.close()\n            \n            self.documents = self._split_text(full_text)\n            print(f\"\u6210\u529f\u63d0\u53d6 {len(self.documents)} \u4e2a\u6587\u672c\u5757\")\n            \n            return self.documents\n            \n        except Exception as e:\n            print(f\"\u52a0\u8f7dPDF\u6587\u4ef6\u5931\u8d25: {e}\")\n            return []\n    \n    def _split_text(self, text: str) -> List[str]:\n        \"\"\"\u5c06\u6587\u672c\u5206\u5272\u6210\u5c0f\u5757\"\"\"\n        sentences = re.split(r'[\u3002\uff01\uff1f\uff1b\\n]', text)\n        chunks = []\n        current_chunk = \"\"\n        \n        for sentence in sentences:\n            sentence = sentence.strip()\n            if not sentence:\n                continue\n                \n            if len(current_chunk) + len(sentence) <= self.chunk_size:\n                current_chunk += sentence + \"\u3002\"\n            else:\n                if current_chunk:\n                    chunks.append(current_chunk.strip())\n                current_chunk = sentence + \"\u3002\"\n        \n        if current_chunk:\n            chunks.append(current_chunk.strip())\n        \n        return chunks\n    \n    def build_embeddings(self):\n        \"\"\"\u4e3a\u6240\u6709\u6587\u6863\u6784\u5efa\u5411\u91cf\u8868\u793a\"\"\"\n        if not self.documents:\n            print(\"\u6ca1\u6709\u6587\u6863\u53ef\u4ee5\u6784\u5efa\u5411\u91cf\")\n            return\n            \n        print(\"\u6b63\u5728\u6784\u5efa\u6587\u6863\u5411\u91cf...\")\n        with torch.inference_mode():\n            self.embeddings = self.embedding_model.encode(self.documents, is_query=False)\n        print(\"\u6587\u6863\u5411\u91cf\u6784\u5efa\u5b8c\u6210\")\n    \n    def hybrid_search(self, query: str, top_k_embedding: int = 10, top_k_final: int = 5) -> List[Tuple[str, float, int]]:\n        \"\"\"\n        \u6df7\u5408\u68c0\u7d22\uff1aEmbedding\u7c97\u7b5b + Reranker\u7cbe\u7b5b\n        Args:\n            query: \u67e5\u8be2\u6587\u672c\n            top_k_embedding: Embedding\u9636\u6bb5\u8fd4\u56de\u7684\u5019\u9009\u6570\u91cf\n            top_k_final: \u6700\u7ec8\u8fd4\u56de\u7684\u7ed3\u679c\u6570\u91cf\n        Returns:\n            \u7ed3\u679c\u5217\u8868\uff0c\u6bcf\u4e2a\u5143\u7d20\u5305\u542b(\u6587\u6863\u5185\u5bb9, \u76f8\u5173\u6027\u5206\u6570, \u6587\u6863\u7d22\u5f15)\n        \"\"\"\n        if self.embeddings is None:\n            print(\"\u8bf7\u5148\u8c03\u7528 build_embeddings() \u6784\u5efa\u6587\u6863\u5411\u91cf\")\n            return []\n        \n        print(f\"\u6b63\u5728\u8fdb\u884c\u6df7\u5408\u68c0\u7d22: {query}\")\n        print(\"=\"*50)\n        \n        # \u7b2c\u4e00\u9636\u6bb5\uff1aEmbedding\u7c97\u7b5b\n        print(\"\u7b2c\u4e00\u9636\u6bb5\uff1aEmbedding\u7c97\u7b5b\")\n        with torch.inference_mode():\n            query_embedding = self.embedding_model.encode([query], is_query=True)\n        \n        similarities = torch.mm(query_embedding, self.embeddings.T)\n        similarities = similarities.squeeze().cpu().numpy()\n        \n        # \u83b7\u53d6top-k\u5019\u9009\n        top_indices = np.argsort(similarities)[::-1][:top_k_embedding]\n        candidates = [(self.documents[idx], similarities[idx], idx) for idx in top_indices]\n        \n        print(f\"Embedding\u9636\u6bb5\u627e\u5230 {len(candidates)} \u4e2a\u5019\u9009\u6587\u6863:\")\n        for i, (doc, score, idx) in enumerate(candidates, 1):\n            print(f\"  \u5019\u9009{i}: \u76f8\u4f3c\u5ea6={score:.4f}, \u7d22\u5f15={idx}\")\n            print(f\"    \u5185\u5bb9: {doc[:80]}...\")\n        \n        # \u7b2c\u4e8c\u9636\u6bb5\uff1aReranker\u7cbe\u7b5b\n        print(f\"\\n\u7b2c\u4e8c\u9636\u6bb5\uff1aReranker\u7cbe\u7b5b\")\n        if len(candidates) > 0:\n            # \u6784\u5efa\u67e5\u8be2-\u6587\u6863\u5bf9\n            pairs = [(query, doc) for doc, _, _ in candidates]\n            \n            # \u4f7f\u7528Reranker\u8ba1\u7b97\u76f8\u5173\u6027\u5206\u6570\n            reranker_scores = self.reranker_model.compute_scores(pairs)\n            \n            # \u7ec4\u5408\u7ed3\u679c\n            reranked_results = []\n            for i, (doc, embedding_score, idx) in enumerate(candidates):\n                reranker_score = reranker_scores[i]\n                reranked_results.append((doc, reranker_score, idx, embedding_score))\n            \n            # \u6309Reranker\u5206\u6570\u6392\u5e8f\n            reranked_results.sort(key=lambda x: x[1], reverse=True)\n            \n            print(f\"Reranker\u9636\u6bb5\u91cd\u65b0\u6392\u5e8f\u7ed3\u679c:\")\n            for i, (doc, reranker_score, idx, embedding_score) in enumerate(reranked_results, 1):\n                print(f\"  \u6392\u5e8f{i}: Reranker\u5206\u6570={reranker_score:.4f}, Embedding\u5206\u6570={embedding_score:.4f}, \u7d22\u5f15={idx}\")\n                print(f\"    \u5185\u5bb9: {doc[:80]}...\")\n            \n            # \u8fd4\u56de\u6700\u7ec8\u7ed3\u679c\n            final_results = [(doc, reranker_score, idx) for doc, reranker_score, idx, _ in reranked_results[:top_k_final]]\n            return final_results\n        \n        return []\n    \n    def embedding_only_search(self, query: str, top_k: int = 5) -> List[Tuple[str, float, int]]:\n        \"\"\"\u4ec5\u4f7f\u7528Embedding\u7684\u641c\u7d22\"\"\"\n        if self.embeddings is None:\n            print(\"\u8bf7\u5148\u8c03\u7528 build_embeddings() \u6784\u5efa\u6587\u6863\u5411\u91cf\")\n            return []\n        \n        print(f\"\u4ec5\u4f7f\u7528Embedding\u641c\u7d22: {query}\")\n        \n        with torch.inference_mode():\n            query_embedding = self.embedding_model.encode([query], is_query=True)\n        \n        similarities = torch.mm(query_embedding, self.embeddings.T)\n        similarities = similarities.squeeze().cpu().numpy()\n        \n        top_indices = np.argsort(similarities)[::-1][:top_k]\n        \n        results = []\n        for idx in top_indices:\n            score = float(similarities[idx])\n            doc_content = self.documents[idx]\n            results.append((doc_content, score, idx))\n        \n        return results\n    \n    def reranker_only_search(self, query: str, top_k: int = 5) -> List[Tuple[str, float, int]]:\n        \"\"\"\u4ec5\u4f7f\u7528Reranker\u7684\u641c\u7d22\uff08\u5bf9\u6240\u6709\u6587\u6863\uff09\"\"\"\n        print(f\"\u4ec5\u4f7f\u7528Reranker\u641c\u7d22: {query}\")\n        \n        if len(self.documents) == 0:\n            print(\"\u6ca1\u6709\u6587\u6863\u53ef\u4ee5\u641c\u7d22\")\n            return []\n        \n        # \u6784\u5efa\u6240\u6709\u67e5\u8be2-\u6587\u6863\u5bf9\n        pairs = [(query, doc) for doc in self.documents]\n        \n        # \u4f7f\u7528Reranker\u8ba1\u7b97\u76f8\u5173\u6027\u5206\u6570\n        reranker_scores = self.reranker_model.compute_scores(pairs)\n        \n        # \u7ec4\u5408\u7ed3\u679c\u5e76\u6392\u5e8f\n        results = []\n        for i, (doc, score) in enumerate(zip(self.documents, reranker_scores)):\n            results.append((doc, score, i))\n        \n        results.sort(key=lambda x: x[1], reverse=True)\n        return results[:top_k]\n\ndef main():\n    # \u521d\u59cb\u5316\u68c0\u7d22\u5668\n    retriever = HybridPDFRetriever()\n    \n    # \u52a0\u8f7dPDF\n    pdf_path = r\"C:\\Users\\sanrome\\Documents\\\u4e09\u90ce\u7b80\u5386-zh.pdf\"\n    documents = retriever.load_pdf(pdf_path)\n    if not documents:\n        print(\"PDF\u52a0\u8f7d\u5931\u8d25\uff0c\u7a0b\u5e8f\u9000\u51fa\")\n        return\n    \n    # \u6784\u5efa\u5411\u91cf\n    retriever.build_embeddings()\n    \n    print(\"\\n\" + \"=\"*80)\n    print(\"=== \u6df7\u5408\u68c0\u7d22\u7cfb\u7edf\u6d4b\u8bd5 ===\")\n    print(\"=\"*80)\n    \n    # \u6d4b\u8bd5\u67e5\u8be2\n    test_queries = [\n        \"\u674e\u660e\u7684\u5de5\u4f5c\u7ecf\u9a8c\",\n        \"\u6280\u672f\u6280\u80fd\u548c\u7f16\u7a0b\u8bed\u8a00\",\n        \"\u5e94\u8058\u7684\u5c97\u4f4d\u548c\u516c\u53f8\",\n        \"\u6d4b\u8bd5\u548c\u5f00\u53d1\u7ecf\u9a8c\"\n    ]\n    \n    for query in test_queries:\n        print(f\"\\n{'='*60}\")\n        print(f\"\u67e5\u8be2: {query}\")\n        print(f\"{'='*60}\")\n        \n        # 1. \u4ec5Embedding\u641c\u7d22\n        print(\"\\n1\ufe0f\u20e3 \u4ec5Embedding\u641c\u7d22:\")\n        embedding_results = retriever.embedding_only_search(query, top_k=3)\n        for i, (doc, score, idx) in enumerate(embedding_results, 1):\n            print(f\"  \u7ed3\u679c{i}: \u76f8\u4f3c\u5ea6={score:.4f}, \u7d22\u5f15={idx}\")\n            print(f\"    \u5185\u5bb9: {doc[:100]}...\")\n        \n        # 2. \u4ec5Reranker\u641c\u7d22\n        print(\"\\n2\ufe0f\u20e3 \u4ec5Reranker\u641c\u7d22:\")\n        reranker_results = retriever.reranker_only_search(query, top_k=3)\n        for i, (doc, score, idx) in enumerate(reranker_results, 1):\n            print(f\"  \u7ed3\u679c{i}: \u76f8\u5173\u6027={score:.4f}, \u7d22\u5f15={idx}\")\n            print(f\"    \u5185\u5bb9: {doc[:100]}...\")\n        \n        # 3. \u6df7\u5408\u641c\u7d22\n        print(\"\\n3\ufe0f\u20e3 \u6df7\u5408\u641c\u7d22 (Embedding + Reranker):\")\n        hybrid_results = retriever.hybrid_search(query, top_k_embedding=5, top_k_final=3)\n        for i, (doc, score, idx) in enumerate(hybrid_results, 1):\n            print(f\"  \u7ed3\u679c{i}: \u6700\u7ec8\u76f8\u5173\u6027={score:.4f}, \u7d22\u5f15={idx}\")\n            print(f\"    \u5185\u5bb9: {doc[:100]}...\")\n\nif __name__ == \"__main__\":\n    main() <\/code><\/pre>\n","protected":false},"excerpt":{"rendered":"
                      \u6700\u8fd1\u51e0\u5929\u963f\u91cc\u53d1\u5e03\u4e86Reranke...<\/p>\n","protected":false},"author":1,"featured_media":2520,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[4],"tags":[],"class_list":["post-2519","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-4"],"_links":{"self":[{"href":"https:\/\/sanlangcode.com\/index.php\/wp-json\/wp\/v2\/posts\/2519","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/sanlangcode.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/sanlangcode.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/sanlangcode.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/sanlangcode.com\/index.php\/wp-json\/wp\/v2\/comments?post=2519"}],"version-history":[{"count":7,"href":"https:\/\/sanlangcode.com\/index.php\/wp-json\/wp\/v2\/posts\/2519\/revisions"}],"predecessor-version":[{"id":2527,"href":"https:\/\/sanlangcode.com\/index.php\/wp-json\/wp\/v2\/posts\/2519\/revisions\/2527"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/sanlangcode.com\/index.php\/wp-json\/wp\/v2\/media\/2520"}],"wp:attachment":[{"href":"https:\/\/sanlangcode.com\/index.php\/wp-json\/wp\/v2\/media?parent=2519"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/sanlangcode.com\/index.php\/wp-json\/wp\/v2\/categories?post=2519"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/sanlangcode.com\/index.php\/wp-json\/wp\/v2\/tags?post=2519"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}