{"id":1260,"date":"2025-07-16T20:38:32","date_gmt":"2025-07-16T12:38:32","guid":{"rendered":"https:\/\/lingbo.online\/?p=1260"},"modified":"2025-11-22T16:10:57","modified_gmt":"2025-11-22T08:10:57","slug":"rag%e9%a1%b9%e7%9b%ae%e6%b4%9e%e5%af%9f","status":"publish","type":"post","link":"https:\/\/lingbo.online\/index.php\/uncategorized\/rag%e9%a1%b9%e7%9b%ae%e6%b4%9e%e5%af%9f\/","title":{"rendered":"RAG\u9879\u76ee\u6d1e\u5bdf"},"content":{"rendered":"<h3>\u9879\u76ee\u5f62\u6001<\/h3>\n<ul>\n<li>\u6587\u4ef6\u89e3\u6790\u73af\u8282\u4e0e\u77e5\u8bc6\u5e93\u7ed1\u5b9a\uff0c\u652f\u6301\u5728\u521b\u5efa\u77e5\u8bc6\u5e93\u65f6\u7ed1\u5b9a\u540c\u4e49\u8bcd\u5e93\uff0c\u4ece\u800c\u5728\u77e5\u8bc6\u5207\u7247\u89e3\u6790\u65f6\u4e3a\u6bcf\u4e2achunk\u914d\u7f6e\u5173\u952e\u8bcd\u8fc7\u6ee4\u4fe1\u606f\uff0c\u51cf\u5c11\u68c0\u7d22\u6210\u672c\uff0c\u63d0\u9ad8\u68c0\u7d22\u6548\u7387\u3002<\/li>\n<li>\u5f53\u524d\u5411\u91cf\u5316\u901f\u5ea6300ms\uff0c\u53ec\u56de\u901f\u5ea61s\u5de6\u53f3\uff0c\u7cbe\u6392\u901f\u5ea63s\u5de6\u53f3\u3002<\/li>\n<\/ul>\n<h4>bce-embedding-base-v1<\/h4>\n<p>BCEmbedding-base-v1\uff08Bilingual and Crosslingual Embedding, BCEmbedding\uff09\u662f\u7f51\u6613\u6709\u9053\u9000\u51fa\u7684\u8de8\u8bed\u79cd\u8bed\u4e49\u68c0\u7d22\u6a21\u578b\uff0c\u4e8e2024\u5e741\u67083\u65e5\u53d1\u5e03\uff0c\u4e3b\u8981\u7528\u4e8e\u68c0\u7d22\u589e\u5f3a\u751f\u6210(RAG)\u4efb\u52a1\uff0c\u662fQAnything\u5f00\u6e90\u9879\u76ee\u7684\u68c0\u7d22\u57fa\u77f3\u3002\u8be5\u6a21\u578b\u7684\u5d4c\u5165\u957f\u5ea6\u4e3a768\uff0c\u6700\u5927\u8f93\u5165\u957f\u5ea6\u4e3a512token\u3002\u4f5c\u4e3aRAG\u6280\u672f\u7ebf\u8def\u4e2d\u6700\u4e3a\u91cd\u8981\u548c\u57fa\u7840\u7684\u4e00\u73af\uff0c\u4e8c\u9636\u6bb5\u68c0\u7d22\u5668\u4e00\u822c\u7531\u53ec\u56de\u548c\u7cbe\u6392\u4e24\u4e2a\u6a21\u5757\u7ec4\u6210\u3002<br \/>\n\u4e8c\u9636\u6bb5\u68c0\u7d22\u5668\u5305\u542b<code>\u53ec\u56de<\/code>\u548c<code>\u7cbe\u6392<\/code>\u4e24\u4e2a\u9636\u6bb5\uff0c\u80fd\u591f\u5f88\u597d\u7684\u5e73\u8861\u68c0\u7d22\u6027\u80fd\u548c\u6548\u7387\u3002<br \/>\n<img decoding=\"async\" src=\"https:\/\/youpaiyun.lingbo.online\/2025\/07\/20250716115527180.jpeg\" alt=\"\" \/><\/p>\n<p>\u53ec\u56de\u9636\u6bb5\u5e38\u91c7\u7528\u57fa\u4e8e\u5411\u91cf\u7684\u5bc6\u96c6\u68c0\u7d22\u65b9\u6cd5\uff0c\u5bf9\u7528\u6237\u95ee\u9898\u548c\u77e5\u8bc6\u5e93\u8bed\u6599\u8fdb\u884c\u5411\u91cf\u5316\u63d0\u53d6\uff0c\u7136\u540e\u641c\u7d22\u548c\u7528\u6237\u95ee\u9898\u8bed\u4e49\u76f8\u8fd1\u7684\u80cc\u666f\u4fe1\u606f\u3002\u63d0\u53d6\u8bed\u4e49\u5411\u91cf\u7684Embedding\u6a21\u578b\u4e00\u822c\u91c7\u7528dual-encoder\u7684\u67b6\u6784\uff0c\u53ef\u4ee5\u79bb\u7ebf(offline)\u5bf9\u5e9e\u5927\u7684\u77e5\u8bc6\u5e93\u8bed\u6599\u63d0\u53d6\u8bed\u4e49\u5411\u91cf\u3002\u7528\u6237\u63d0\u95ee\u65f6\uff0c\u6a21\u578b\u53ea\u9700\u8981\u5b9e\u65f6\u7684\u63d0\u53d6\u7528\u6237\u95ee\u9898\u7684\u8bed\u4e49\u5411\u91cf\uff0c\u5c31\u53ef\u4ee5\u5229\u7528\u5411\u91cf\u6570\u636e\u5e93\u8fdb\u884c\u8bed\u4e49\u68c0\u7d22\u3002\u5728\u8fd9\u4e2a\u8fc7\u7a0b\u4e2d\uff0c\u5927\u89c4\u6a21\u77e5\u8bc6\u5e93\u8bed\u6599\u7684\u8bed\u4e49\u5411\u91cf\u63d0\u53d6\u662f\u79bb\u7ebf\u3001\u9759\u6001\u7684\u8fc7\u7a0b\uff0c\u6a21\u578b\u5728\u63d0\u53d6\u7528\u6237\u95ee\u9898\u548c\u77e5\u8bc6\u5e93\u8bed\u6599\u7684\u8bed\u4e49\u5411\u91cf\u65f6\uff0c\u6ca1\u6709\u4fe1\u606f\u4ea4\u4e92\u3002\u8be5\u65b9\u5f0f\u7684\u597d\u5904\u662f\u68c0\u7d22\u6548\u7387\u975e\u5e38\u9ad8\uff0c\u4f46\u662f\u8fd9\u4e5f\u9650\u5236\u4e86\u8bed\u4e49\u68c0\u7d22\u6027\u80fd\u7684\u4e0a\u9650\u3002<br \/>\n\u7cbe\u6392\u9636\u6bb5\u662f\u4e3a\u4e86\u89e3\u51b3\u4fe1\u606f\u4ea4\u4e92\u7684\u95ee\u9898\uff0c\u91c7\u7528cross-encoder\u67b6\u6784\u3002Reranker\u6a21\u578b\u53ef\u4ee5\u5b9e\u73b0\u7528\u6237\u95ee\u9898\u548c\u77e5\u8bc6\u5e93\u8bed\u6599\u7684\u4fe1\u606f\u4ea4\u4e92\uff0c\u4f7f\u6a21\u578b\u80fd\u591f\u201c\u89c1\u673a\u884c\u4e8b\u201d\u7684\u8bc6\u522b\u5230\u66f4\u52a0\u51c6\u786e\u7684\u8bed\u4e49\u5173\u7cfb\uff0c\u7b97\u6cd5\u6027\u80fd\u4e0a\u9650\u53ef\u4ee5\u5f88\u9ad8\u3002\u8be5\u65b9\u5f0f\u7684\u7f3a\u70b9\u662f\uff0c\u9700\u8981\u5bf9\u7528\u6237\u95ee\u9898\u548c\u77e5\u8bc6\u5e93\u8bed\u6599\u8fdb\u884c\u5728\u7ebf\uff08online\uff09\u7684\u8bed\u4e49\u5173\u7cfb\u63d0\u53d6\uff0c\u6548\u7387\u8f83\u4f4e\uff0c\u65e0\u6cd5\u5bf9\u5168\u91cf\u7684\u77e5\u8bc6\u5e93\u8bed\u6599\u8fdb\u884c\u5b9e\u65f6\u5904\u7406\u3002<br \/>\n\u7ed3\u5408\u53ec\u56de\u548c\u7cbe\u6392\u7684\u4f18\u52bf\uff0c\u5728\u53ec\u56de\u9636\u6bb5\u53ef\u4ee5\u5feb\u901f\u627e\u5230\u7528\u6237\u95ee\u9898\u76f8\u5173\u7684\u6587\u672c\u7247\u6bb5\uff0c\u7cbe\u6392\u9636\u6bb5\u53ef\u4ee5\u5c06\u6b63\u786e\u7684\u76f8\u5173\u7247\u6bb5\u5c3d\u53ef\u80fd\u6392\u5728\u9760\u524d\u4f4d\u7f6e\uff0c\u5e76\u8fc7\u6ee4\u6389\u4f4e\u8d28\u91cf\u7684\u7247\u6bb5\u3002\u4e8c\u9636\u6bb5\u68c0\u7d22\u53ef\u4ee5\u5f88\u597d\u7684\u6743\u8861\u68c0\u7d22\u6548\u679c\u548c\u6548\u7387\uff0c\u5177\u6709\u5de8\u5927\u7684\u5e94\u7528\u4ef7\u503c\u3002<br \/>\n<img decoding=\"async\" src=\"https:\/\/youpaiyun.lingbo.online\/2025\/07\/20250716115757725.jpeg\" alt=\"\" \/><\/p>\n<h4>bge-large-zh-v1.5<\/h4>\n<p>Bge\uff08BAAI General Embedding\uff09-large-zh-v1.5\u662f\u7531BAAI\uff08Beijing Academy of Artificial Intelligence\uff09\u5f00\u53d1\u7684\u4e2d\u6587\u5d4c\u5165\u6a21\u578b\u3002\u5d4c\u5165\u957f\u5ea6\u4e3a1024\uff0c\u6700\u5927\u8f93\u5165\u957f\u5ea6\u4e3a512\u3002\u8be5\u6a21\u578b\u5728\u4e2d\u6587\u8bed\u5883\u4e0b\u7684\u4efb\u52a1\u4e2d\u8868\u73b0\u51fa\u8272\uff0c\u5982\u53e5\u5b50\u76f8\u4f3c\u5ea6\u3001\u6587\u672c\u68c0\u7d22\u548c\u95ee\u7b54\u7cfb\u7edf\uff0c\u5177\u6709\u51c6\u786e\u7387\u9ad8\u3001\u63a8\u7406\u901f\u5ea6\u5feb\u3001\u8d44\u6e90\u6d88\u8017\u4f4e\u7b49\u4f18\u52bf\u3002\u652f\u6301512\u4e2atoken\u7684\u8f93\u5165\uff0c\u9002\u7528\u4e8e\u591a\u8bed\u8a00\u73af\u5883\u4e0b\u7684\u6587\u672c\u68c0\u7d22\u4efb\u52a1\u3002<\/p>\n<h4>bge-m3<\/h4>\n<p>BGE-M3\uff08Multi-language, Multi-function, Multi-granularity\uff08\u591a\u8bed\u8a00\u3001\u591a\u529f\u80fd\u3001\u591a\u7c92\u5ea6\uff09\uff09\u662f\u667a\u6e90\u63a8\u51fa\u7684\u65b0\u4e00\u4ee3\u901a\u7528\u8bed\u4e49\u5411\u91cf\u6a21\u578b\uff0c\u652f\u6301100\u591a\u79cd\u8bed\u8a00\uff0c\u5177\u5907\u9ad8\u6548\u7684\u591a\u8bed\u8a00\u68c0\u7d22\u80fd\u529b\u3002\u5b83\u80fd\u5904\u7406\u4e0d\u540c\u7c92\u5ea6\u7684\u6587\u672c\uff0c\u5d4c\u5165\u957f\u5ea6\u4e3a1024\u7ef4\uff0c\u6700\u5927\u8f93\u5165\u957f\u5ea6\u4e3a8192tokens\uff0c\u96c6\u6210\u4e86\u591a\u79cd\u68c0\u7d22\u529f\u80fd\uff0c\u6027\u80fd\u4f18\u5f02\u3002<\/p>\n<h3>Reranker\u8bad\u7ec3<\/h3>\n<h4>Qwen3-Rerank<\/h4>\n<p>Qwen3-Rerank\u7cfb\u7c7b\u5305\u62ec0.6B\u30014B\u548c8B\u4e09\u79cd\u7c7b\u578b\u3002\u5bf9\u4e8eqwen-reranker\uff0c\u4ed6\u7684\u8f93\u5165\u662f\u3010\u67e5\u8be2-\u6587\u6863\u3011\u4e8c\u5143\u7ec4\uff0c\u56fa\u5b9a\u8f93\u51fa\u4e3ayes\/no\uff0c\u56e0\u6b64\u8bad\u7ec3\u6837\u672c\u7684\u683c\u5f0f\u7c7b\u4f3c\u4e8e\uff1a<\/p>\n<pre><code class=\"language-json\">{\n&quot;query&quot;:XXX,\n&quot;doc&quot;:XXX\uff0c\n&quot;label&quot;:yes\/no\n}<\/code><\/pre>\n<p>\u5176\u4e2d\uff0c\u6b63\u6837\u672c\u4e3a\u9886\u57df\u5185\u771f\u5b9e\u76f8\u5173\u95ee\u7b54\/\u6bb5\u843d\uff1b\u8d1f\u6837\u672c\u53ef\u4ee5\u662f\u968f\u673a\u62bd\u53d6\u7684\u4e0d\u76f8\u5173\u6bb5\u843d\uff0c\u6216\u8005\u662f\u4f7f\u7528\u96be\u8d1f\u6837\u672c\u6316\u6398\u7b56\u7565\u627e\u5230\u7684\u8d1f\u6837\u672c\u3002\u6b63\u8d1f\u6837\u672c\u8bbe\u7f6e1:1\u5de6\u53f3\u5373\u53ef\uff0c\u51e0\u5343\u6761\u5c31\u80fd\u89c1\u6548\u3002<\/p>\n<p>\u5728\u5b9e\u8df5\u4e2d\uff0c\u5229\u7528markdown\u8bed\u6599+\u539f\u59cb\u95ee\u7b54\uff0c\u6784\u5efa\u95ee\u9898-\u6b63\u4f8b-\u8d1f\u4f8b\u6837\u672c\u5bf9\u3002\u5176\u4e2d\u8d1f\u4f8b\u7684\u83b7\u53d6\u65b9\u6cd5\u662f\uff1a<\/p>\n<ul>\n<li>\n<p>\u9996\u5148\u4f7f\u7528\u5411\u91cf\u53ec\u56de\u6700\u96be\u7247\u6bb5\uff0c\u901a\u8fc7faiss-ip\u641c\u7d22\u505a\u53cd\u5411\u641c\u7d22\uff0c\u5f97\u5230\u6700\u4e0d\u76f8\u5173\u7684\u7247\u6bb5\u4f5c\u4e3a\u5019\u9009\u6c60<\/p>\n<\/li>\n<li>\n<p>\u4ece\u540e\u9009\u7247\u6bb5\u4e2d\u8fdb\u884c\u968f\u673a\u62bd\u53d630\u6761\uff0c\u5e76\u4f7f\u7528qwen-rerank-8B\u8fdb\u884c\u6253\u5206\uff0c\u4fdd\u755930\u6761\u8d1f\u4f8b\u53ca\u5176\u91cd\u6392\u5206\u6570<\/p>\n<\/li>\n<li>\n<p>\u6700\u540e\u5f97\u52302w+\u6837\u672c\u5bf9\uff0c\u4ece\u800c\u8fdb\u884crerank\u5fae\u8c03\u6216\u5343\u95eererank\u5fae\u8c03<\/p>\n<h4>\u77e5\u8bc6\u8865\u5145<\/h4>\n<\/li>\n<li>\n<p>\u5d4c\u5165\u957f\u5ea6\uff1a\u6a21\u578b\u5c06\u6bcf\u4e2a\u8bcd\u6216Token\u8f6c\u6362\u6210\u7684\u5411\u91cf\u7684\u7ef4\u5ea6\uff08\u5373\u5411\u91cf\u957f\u5ea6\uff09\u3002\u8be5\u503c\u51b3\u5b9a\u6a21\u578b\u8868\u793a\u8bcd\u6c47\u8bed\u4e49\u7279\u5f81\u7684\u4e30\u5bcc\u7a0b\u5ea6\u3002\u7ef4\u5ea6\u8d8a\u9ad8\uff0c\u5219\u8868\u8fbe\u80fd\u529b\u8d8a\u5f3a\uff0c\u4f46\u662f\u8ba1\u7b97\u6210\u672c\u4e5f\u66f4\u9ad8\u3002<\/p>\n<\/li>\n<li>\n<p>\u6700\u5927\u8f93\u5165\u957f\u5ea6\uff1a\u6a21\u578b\u5355\u8bcd\u80fd\u5904\u7406\u7684\u6700\u5927Token\u5e8f\u5217\u957f\u5ea6\uff08\u5305\u62ec\u6587\u672c\u548c\u7279\u6b8a\u6807\u8bb0\u5982[CLS]\u3001[SEP]\uff09\u3002\u53d7\u6a21\u578b\u7684\u4f4d\u7f6e\u7f16\u7801\uff08Positional Encoding\uff09\u548c\u5185\u5b58\u9650\u5236\u5f71\u54cd\u3002\u8d85\u51fa\u90e8\u5206\u4f1a\u88ab\u622a\u65ad\u6216\u8005\u62d2\u7edd\u3002<\/p>\n<\/li>\n<li>\n<p>[CLS]\uff1aClassification Token\uff0c\u5728\u6a21\u578b\u8f93\u5165\u7684\u5f00\u5934\u63d2\u5165[CLS],\u5176\u5bf9\u5e94\u7684\u8f93\u51fa\u5411\u91cf\u88ab\u7528\u4f5c\u6574\u4e2a\u8f93\u5165\u5e8f\u5217\u7684\u201c\u8bed\u4e49\u6458\u8981\u201d,\u5c24\u5176\u7528\u4e8e\u5206\u7c7b\u4efb\u52a1\uff08\u5982\u60c5\u611f\u5206\u6790\u3001\u6587\u672c\u5339\u914d\u3001\u53e5\u5b50\u5206\u7c7b\u7b49\uff09<\/p>\n<\/li>\n<li>\n<p>[SEP]:Separator Token\uff0c\u7528\u4e8e\u5206\u9694\u4e24\u4e2a\u4e0d\u540c\u7684\u53e5\u5b50\u6216\u7247\u6bb5\uff0c\u5e2e\u52a9\u6a21\u578b\u533a\u5206\u4ed6\u4eec\u7684\u8fb9\u754c\u3002\u5e38\u89c1\u4e8e\u53e5\u5b50\u5bf9\u4efb\u52a1\uff08\u95ee\u7b54\u3001\u81ea\u7136\u8bed\u8a00\u63a8\u7406\u3001\u53e5\u5b50\u76f8\u4f3c\u5ea6\uff09\u6216\u5355\u53e5\u4efb\u52a1\uff0c\u5373\u5728\u53e5\u5b50\u672b\u5c3e\u6dfb\u52a0[SEP]\u8868\u793a\u53e5\u5b50\u7ed3\u675f\u3002<\/p>\n<\/li>\n<\/ul>\n<h4>\u9057\u7559\u9879<\/h4>\n<ul>\n<li>\u53ec\u56de\u901f\u5ea6\uff0c\u7cbe\u6392\u901f\u5ea6<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>\u9879\u76ee\u5f62\u6001 \u6587\u4ef6\u89e3\u6790\u73af\u8282\u4e0e\u77e5\u8bc6\u5e93\u7ed1\u5b9a\uff0c\u652f\u6301\u5728\u521b\u5efa\u77e5\u8bc6\u5e93\u65f6\u7ed1\u5b9a\u540c\u4e49\u8bcd\u5e93\uff0c\u4ece\u800c\u5728\u77e5\u8bc6\u5207\u7247\u89e3\u6790\u65f6\u4e3a\u6bcf\u4e2achunk\u914d\u7f6e\u5173\u952e\u8bcd\u8fc7\u6ee4\u4fe1\u606f\uff0c\u51cf\u5c11\u68c0 &#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"emotion":"","emotion_color":"","title_style":"","license":"","footnotes":""},"categories":[1],"tags":[],"class_list":["post-1260","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/posts\/1260","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/comments?post=1260"}],"version-history":[{"count":6,"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/posts\/1260\/revisions"}],"predecessor-version":[{"id":1428,"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/posts\/1260\/revisions\/1428"}],"wp:attachment":[{"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/media?parent=1260"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/categories?post=1260"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/lingbo.online\/index.php\/wp-json\/wp\/v2\/tags?post=1260"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}