
{"id":2133,"date":"2024-05-05T00:44:15","date_gmt":"2024-05-04T16:44:15","guid":{"rendered":"https:\/\/infernews.com\/?p=2133"},"modified":"2024-05-05T03:04:00","modified_gmt":"2024-05-04T19:04:00","slug":"2404-05961-llm2vec-large-language-models-are-secretly-powerful-text-encoders","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/2404-05961-llm2vec-large-language-models-are-secretly-powerful-text-encoders\/","title":{"rendered":"[2404.05961] LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders"},"content":{"rendered":"<p><span style=\"background-color: rgba(5, 69, 37, 0.03); color: rgba(0, 8, 5, 0.898); font-family: 'Open Sans', sans-serif; font-size: 15.75px;\">LLM2Vec\uff0c\u9019\u662f\u4e00\u7a2e\u5c07\u50c5\u89e3\u78bc\u5668\u7684\u5927\u578b\u8a9e\u8a00\u6a21\u578b \uff08LLM\uff09 \u8f49\u63db\u70ba\u529f\u80fd\u5f37\u5927\u7684\u6587\u672c\u7de8\u78bc\u5668\u7684\u65b9\u6cd5\u3002LLM2Vec \u7531\u4e09\u500b\u6b65\u9a5f\u7d44\u6210\uff1a\u555f\u7528\u96d9\u5411\u6ce8\u610f\u529b\u3001\u906e\u7f69\u4e0b\u4e00\u500b\u4ee4\u724c\u9810\u6e2c\u548c\u7121\u76e3\u7763\u5c0d\u6bd4\u5b78\u7fd2\u3002\u901a\u904e\u5c07 LLM2Vec \u61c9\u7528\u65bc\u4e09\u500b\u6d41\u884c\u7684 LLM \u4e26\u5728\u55ae\u8a5e\u548c\u5e8f\u5217\u7d1a\u4efb\u52d9\u4e0a\u8a55\u4f30\u5b83\u5011\uff0c\u8a72\u7814\u7a76\u5c55\u793a\u4e86\u8207\u50c5\u7de8\u78bc\u5668\u6a21\u578b\u76f8\u6bd4\u7684\u5353\u8d8a\u6027\u80fd\uff0c\u4e26\u5728\u6d77\u91cf\u6587\u672c\u5d4c\u5165\u57fa\u6e96 \uff08MTEB\uff09 \u4e0a\u5be6\u73fe\u4e86\u65b0\u7684\u6700\u5148\u9032\u6c34\u6e96\u3002\u4e3b\u8981\u767c\u73fe\u5305\u62ec\u5728\u55ae\u8a5e\u7d1a\u4efb\u52d9\u4e0a\u512a\u65bc\u50c5\u7de8\u78bc\u5668\u6a21\u578b\uff0c\u5728MTEB\u4e0a\u9054\u5230\u65b0\u7684\u7121\u76e3\u7763\u72c0\u614b\uff0c\u4ee5\u53ca\u901a\u904e\u76e3\u7763\u5c0d\u6bd4\u5b78\u7fd2\u5be6\u73fe\u6700\u5148\u9032\u7684\u6027\u80fd\u3002\u8a72\u7814\u7a76\u5f37\u8abf\u4e86\u50c5\u89e3\u78bc\u5668 LLM \u4f5c\u70ba\u901a\u7528\u6587\u672c\u7de8\u78bc\u5668\u7684\u6f5b\u529b\uff0c\u53ea\u9700\u6975\u5c11\u7684\u9069\u61c9\u3002LLM2Vec \u65b9\u6cd5\u5177\u6709\u53c3\u6578\u6548\u7387\uff0c\u4e0d\u4f9d\u8cf4\u65bc\u6a19\u8a18\u6578\u64da\uff0c\u4e26\u4e14\u5728\u751f\u6210\u8c50\u5bcc\u7684\u4e0a\u4e0b\u6587\u5316\u6a19\u8a18\u8868\u793a\u65b9\u9762\u975e\u5e38\u6709\u6548\u3002<\/span><\/p>\n<ul style=\"border: 0px solid #e5e7eb; --tw-border-spacing-x: 0; --tw-border-spacing-y: 0; --tw-translate-x: 0; --tw-translate-y: 0; --tw-rotate: 0; --tw-skew-x: 0; --tw-skew-y: 0; --tw-scale-x: 1; --tw-scale-y: 1; --tw-scroll-snap-strictness: proximity; --tw-ring-offset-width: 0px; --tw-ring-offset-color: #fff; --tw-ring-color: rgba(59,130,246,.5); --tw-ring-offset-shadow: 0 0 #0000; --tw-ring-shadow: 0 0 #0000; --tw-shadow: 0 0 #0000; --tw-shadow-colored: 0 0 #0000; margin-bottom: 0px; margin-left: 0px; color: rgba(0, 8, 5, 0.898); font-family: 'Open Sans', sans-serif; font-size: 15.75px; background-color: rgba(5, 69, 37, 0.03); list-style-position: inside !important; list-style-type: initial !important;\">\n<li style=\"box-sizing: border-box; border: 0px solid #e5e7eb; --tw-border-spacing-x: 0; --tw-border-spacing-y: 0; --tw-translate-x: 0; --tw-translate-y: 0; --tw-rotate: 0; --tw-skew-x: 0; --tw-skew-y: 0; --tw-scale-x: 1; --tw-scale-y: 1; --tw-scroll-snap-strictness: proximity; --tw-ring-offset-width: 0px; --tw-ring-offset-color: #fff; --tw-ring-color: rgba(59,130,246,.5); --tw-ring-offset-shadow: 0 0 #0000; --tw-ring-shadow: 0 0 #0000; --tw-shadow: 0 0 #0000; --tw-shadow-colored: 0 0 #0000;\">\u5927\u578b\u7d14\u89e3\u78bc\u5668\u8a9e\u8a00\u6a21\u578b \uff08LLM\uff09 \u662f NLP \u4efb\u52d9\u4e2d\u6700\u5148\u9032\u7684\u7531\u65bc\u56e0\u679c\u6ce8\u610f\u529b\u9650\u5236\uff0cLLM \u5728\u6587\u672c\u5d4c\u5165\u4efb\u52d9\u4e2d\u7684\u63a1\u7528\u901f\u5ea6\u7de9\u6162\u5f15\u5165 LLM2Vec \u5c07\u50c5\u89e3\u78bc\u5668\u7684 LLM \u8f49\u63db\u70ba\u6587\u5b57\u7de8\u78bc\u5668LLM2Vec \u5305\u62ec\u555f\u7528\u96d9\u5411\u6ce8\u610f\u529b\u3001\u906e\u7f69\u4e0b\u4e00\u500b\u4ee4\u724c\u9810\u6e2c\u548c\u7121\u76e3\u7763\u5c0d\u6bd4\u5b78\u7fd2LLM2Vec \u8f49\u63db\u6a21\u578b\u5728\u5b57\u7d1a\u548c\u5e8f\u5217\u7d1a\u4efb\u52d9\u4e0a\u7684\u5353\u8d8a\u6027\u80fd\u5728\u6d77\u91cf\u6587\u672c\u5d4c\u5165\u57fa\u6e96 \uff08MTEB\uff09 \u4e0a\u5be6\u73fe\u65b0\u7684\u6700\u5148\u9032\u7684\u6280\u8853\u53ea\u9700\u6975\u5c11\u7684\u8abf\u6574\u5373\u53ef\u63ed\u793a\u50c5\u89e3\u78bc\u5668 LLM \u4f5c\u70ba\u901a\u7528\u6587\u672c\u7de8\u78bc\u5668\u7684\u529f\u80fd<\/li>\n<li>\u00a0<\/li>\n<li>\u4f7f\u7528LLM2Vec\u5c07\u4efb\u4f55\u50c5\u89e3\u78bc\u5668\u7684\u5927\u578b\u8a9e\u8a00\u6a21\u578b\uff08LLMs\uff09\u8f49\u63db\u70ba\u5f37\u5927\u7684\u6587\u672c\u7de8\u78bc\u5668\u7684\u65b9\u6cd5\u5305\u62ec\u4e09\u500b\u7c21\u55ae\u6b65\u9a5f\uff1a<a>1<\/a>\u3002\u9996\u5148\u662f\u555f\u7528\u96d9\u5411\u6ce8\u610f\u529b\uff0c\u5176\u6b21\u662f\u906e\u7f69\u4e0b\u4e00\u500b\u4ee4\u724c\u9810\u6e2c\uff0c\u6700\u5f8c\u662f\u7121\u76e3\u7763\u5c0d\u6bd4\u5b78\u7fd2\u3002LLM2Vec\u7684\u61c9\u7528\u65bc\u4e09\u500b\u6d41\u884c\u7684LLMs\uff0c\u4e26\u5728\u82f1\u6587\u55ae\u8a5e\u548c\u5e8f\u5217\u7d1a\u4efb\u52d9\u4e0a\u8a55\u4f30\u8f49\u63db\u5f8c\u7684\u6a21\u578b\uff0c\u8868\u73fe\u512a\u65bc\u50c5\u7de8\u78bc\u5668\u6a21\u578b\uff0c\u4e26\u5728\u5927\u6587\u672c\u5d4c\u5165\u57fa\u6e96\uff08MTEB\uff09\u4e0a\u5be6\u73fe\u4e86\u65b0\u7684\u7121\u76e3\u7763\u6700\u5148\u9032\u8868\u73fe\u3002\u6b64\u5916\uff0c\u7d50\u5408LLM2Vec\u548c\u76e3\u7763\u5c0d\u6bd4\u5b78\u7fd2\uff0c\u6211\u5011\u5728\u50c5\u5728\u516c\u958b\u53ef\u7528\u6578\u64da\u4e0a\u8a13\u7df4\u7684\u6a21\u578b\u4e2d\u5be6\u73fe\u4e86MTEB\u7684\u6700\u65b0\u6700\u5148\u9032\u8868\u73fe\u3002<a>13<\/a>\u3002LLM2Vec\u4e0d\u9700\u8981\u6a19\u8a18\u6578\u64da\uff0c\u4e26\u4e14\u5728\u6578\u64da\u548c\u53c3\u6578\u6548\u7387\u65b9\u9762\u975e\u5e38\u6709\u6548\u3002<\/li>\n<\/ul>\n<p><a href=\"https:\/\/arxiv.org\/abs\/2404.05961\" target=\"_blank\" rel=\"noopener\">Paper Link<\/a><\/p>\n<p><a href=\"https:\/\/github.com\/McGill-NLP\/llm2vec\" target=\"_blank\" rel=\"noopener\">GitHub Link<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>LLM2Vec\uff0c\u9019\u662f\u4e00\u7a2e\u5c07\u50c5\u89e3\u78bc\u5668\u7684\u5927\u578b\u8a9e\u8a00\u6a21\u578b \uff08LLM\uff09 \u8f49\u63db\u70ba\u529f\u80fd\u5f37\u5927\u7684\u6587\u672c\u7de8\u78bc\u5668\u7684\u65b9\u6cd5\u3002LLM2Vec \u7531\u4e09\u500b\u6b65\u9a5f\u7d44\u6210\uff1a\u555f\u7528\u96d9\u5411\u6ce8\u610f\u529b\u3001\u906e\u7f69\u4e0b\u4e00\u500b\u4ee4\u724c\u9810\u6e2c\u548c\u7121\u76e3\u7763\u5c0d\u6bd4\u5b78\u7fd2\u3002\u901a\u904e\u5c07 LLM2Vec \u61c9\u7528\u65bc\u4e09\u500b\u6d41\u884c\u7684 LLM \u4e26\u5728\u55ae\u8a5e\u548c\u5e8f\u5217\u7d1a\u4efb\u52d9\u4e0a\u8a55\u4f30\u5b83\u5011\uff0c\u8a72\u7814\u7a76\u5c55\u793a\u4e86\u8207\u50c5\u7de8\u78bc\u5668\u6a21\u578b\u76f8\u6bd4\u7684\u5353\u8d8a\u6027\u80fd\uff0c\u4e26\u5728\u6d77\u91cf\u6587\u672c\u5d4c\u5165\u57fa\u6e96 \uff08MTEB\uff09 \u4e0a\u5be6\u73fe\u4e86\u65b0\u7684\u6700\u5148\u9032\u6c34\u6e96\u3002\u4e3b\u8981\u767c\u73fe\u5305\u62ec\u5728\u55ae\u8a5e\u7d1a\u4efb\u52d9\u4e0a\u512a\u65bc\u50c5\u7de8\u78bc\u5668\u6a21\u578b\uff0c\u5728MTEB\u4e0a\u9054\u5230\u65b0\u7684\u7121\u76e3\u7763\u72c0\u614b\uff0c\u4ee5\u53ca\u901a\u904e\u76e3\u7763\u5c0d\u6bd4\u5b78\u7fd2\u5be6\u73fe\u6700\u5148\u9032\u7684\u6027\u80fd\u3002\u8a72\u7814\u7a76\u5f37\u8abf\u4e86\u50c5\u89e3\u78bc\u5668 LLM \u4f5c\u70ba\u901a\u7528\u6587\u672c\u7de8\u78bc\u5668\u7684\u6f5b\u529b\uff0c\u53ea\u9700\u6975\u5c11\u7684\u9069\u61c9\u3002LLM2Vec \u65b9\u6cd5\u5177\u6709\u53c3\u6578\u6548\u7387\uff0c\u4e0d\u4f9d\u8cf4\u65bc\u6a19\u8a18\u6578\u64da\uff0c\u4e26\u4e14\u5728\u751f\u6210\u8c50\u5bcc\u7684\u4e0a\u4e0b\u6587\u5316\u6a19\u8a18\u8868\u793a\u65b9\u9762\u975e\u5e38\u6709\u6548\u3002 \u5927\u578b\u7d14\u89e3\u78bc\u5668\u8a9e\u8a00\u6a21\u578b \uff08LLM\uff09 \u662f NLP \u4efb\u52d9\u4e2d\u6700\u5148\u9032\u7684\u7531\u65bc\u56e0\u679c\u6ce8\u610f\u529b\u9650\u5236\uff0cLLM \u5728\u6587\u672c\u5d4c\u5165\u4efb\u52d9\u4e2d\u7684\u63a1\u7528\u901f\u5ea6\u7de9\u6162\u5f15\u5165 LLM2Vec \u5c07\u50c5\u89e3\u78bc\u5668\u7684 LLM \u8f49\u63db\u70ba\u6587\u5b57\u7de8\u78bc\u5668LLM2Vec \u5305\u62ec\u555f\u7528\u96d9\u5411\u6ce8\u610f\u529b\u3001\u906e\u7f69\u4e0b\u4e00\u500b\u4ee4\u724c\u9810\u6e2c\u548c\u7121\u76e3\u7763\u5c0d\u6bd4\u5b78\u7fd2LLM2Vec \u8f49\u63db\u6a21\u578b\u5728\u5b57\u7d1a\u548c\u5e8f\u5217\u7d1a\u4efb\u52d9\u4e0a\u7684\u5353\u8d8a\u6027\u80fd\u5728\u6d77\u91cf\u6587\u672c\u5d4c\u5165\u57fa\u6e96 \uff08MTEB\uff09 \u4e0a\u5be6\u73fe\u65b0\u7684\u6700\u5148\u9032\u7684\u6280\u8853\u53ea\u9700\u6975\u5c11\u7684\u8abf\u6574\u5373\u53ef\u63ed\u793a\u50c5\u89e3\u78bc\u5668 LLM \u4f5c\u70ba\u901a\u7528\u6587\u672c\u7de8\u78bc\u5668\u7684\u529f\u80fd \u00a0 \u4f7f\u7528LLM2Vec\u5c07\u4efb\u4f55\u50c5\u89e3\u78bc\u5668\u7684\u5927\u578b\u8a9e\u8a00\u6a21\u578b\uff08LLMs\uff09\u8f49\u63db\u70ba\u5f37\u5927\u7684\u6587\u672c\u7de8\u78bc\u5668\u7684\u65b9\u6cd5\u5305\u62ec\u4e09\u500b\u7c21\u55ae\u6b65\u9a5f\uff1a1\u3002\u9996\u5148\u662f\u555f\u7528\u96d9\u5411\u6ce8\u610f\u529b\uff0c\u5176\u6b21\u662f\u906e\u7f69\u4e0b\u4e00\u500b\u4ee4\u724c\u9810\u6e2c\uff0c\u6700\u5f8c\u662f\u7121\u76e3\u7763\u5c0d\u6bd4\u5b78\u7fd2\u3002LLM2Vec\u7684\u61c9\u7528\u65bc\u4e09\u500b\u6d41\u884c\u7684LLMs\uff0c\u4e26\u5728\u82f1\u6587\u55ae\u8a5e\u548c\u5e8f\u5217\u7d1a\u4efb\u52d9\u4e0a\u8a55\u4f30\u8f49\u63db\u5f8c\u7684\u6a21\u578b\uff0c\u8868\u73fe\u512a\u65bc\u50c5\u7de8\u78bc\u5668\u6a21\u578b\uff0c\u4e26\u5728\u5927\u6587\u672c\u5d4c\u5165\u57fa\u6e96\uff08MTEB\uff09\u4e0a\u5be6\u73fe\u4e86\u65b0\u7684\u7121\u76e3\u7763\u6700\u5148\u9032\u8868\u73fe\u3002\u6b64\u5916\uff0c\u7d50\u5408LLM2Vec\u548c\u76e3\u7763\u5c0d\u6bd4\u5b78\u7fd2\uff0c\u6211\u5011\u5728\u50c5\u5728\u516c\u958b\u53ef\u7528\u6578\u64da\u4e0a\u8a13\u7df4\u7684\u6a21\u578b\u4e2d\u5be6\u73fe\u4e86MTEB\u7684\u6700\u65b0\u6700\u5148\u9032\u8868\u73fe\u300213\u3002LLM2Vec\u4e0d\u9700\u8981\u6a19\u8a18\u6578\u64da\uff0c\u4e26\u4e14\u5728\u6578\u64da\u548c\u53c3\u6578\u6548\u7387\u65b9\u9762\u975e\u5e38\u6709\u6548\u3002 Paper Link GitHub Link<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"googlesitekit_rrm_CAowvqSiDA:productID":"","footnotes":""},"categories":[23],"tags":[],"class_list":["post-2133","post","type-post","status-publish","format-standard","hentry","category-news"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/2133","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=2133"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/2133\/revisions"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=2133"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=2133"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=2133"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}