{"id":36227,"date":"2025-01-08T00:00:22","date_gmt":"2025-01-07T16:00:22","guid":{"rendered":"https:\/\/17aitech.com\/?p=36227"},"modified":"2025-01-08T00:00:22","modified_gmt":"2025-01-07T16:00:22","slug":"%e8%bf%98%e6%98%af%e5%8e%9f%e8%a3%85transformer%e5%a5%bd%ef%bc%81%e5%8c%97%e5%a4%a7%e6%b8%85%e5%8d%8e%e5%9b%a2%e9%98%9f%e5%90%8c%e6%97%b6%e6%8f%ad%e7%a4%bamamba%e7%ad%89%e6%8e%a8%e7%90%86%e7%9f%ad","status":"publish","type":"post","link":"https:\/\/17aitech.com\/?p=36227","title":{"rendered":"\u8fd8\u662f\u539f\u88c5Transformer\u597d\uff01\u5317\u5927\u6e05\u534e\u56e2\u961f\u540c\u65f6\u63ed\u793aMamba\u7b49\u63a8\u7406\u77ed\u677f"},"content":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-10-21\" target=\"_blank\">\u8fd8\u662f\u539f\u88c5Transformer\u597d\uff01\u5317\u5927\u6e05\u534e\u56e2\u961f\u540c\u65f6\u63ed\u793aMamba\u7b49\u63a8\u7406\u77ed\u677f<\/a><\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-43abb2d37958289a28016caa2c553119.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-43abb2d37958289a28016caa2c553119.png\"><\/a><\/section>\n<blockquote data-author-name=\"\" data-content-utf8-length=\"166\" data-source-title=\"\" data-type=\"2\" data-url=\"\">\n<section>\n<section>AIxiv\u4e13\u680f\u662f\u673a\u5668\u4e4b\u5fc3\u53d1\u5e03\u5b66\u672f\u3001\u6280\u672f\u5185\u5bb9\u7684\u680f\u76ee\u3002\u8fc7\u53bb\u6570\u5e74\uff0c\u673a\u5668\u4e4b\u5fc3AIxiv\u4e13\u680f\u63a5\u6536\u62a5\u9053\u4e862000\u591a\u7bc7\u5185\u5bb9\uff0c\u8986\u76d6\u5168\u7403\u5404\u5927\u9ad8\u6821\u4e0e\u4f01\u4e1a\u7684\u9876\u7ea7\u5b9e\u9a8c\u5ba4\uff0c\u6709\u6548\u4fc3\u8fdb\u4e86\u5b66\u672f\u4ea4\u6d41\u4e0e\u4f20\u64ad\u3002\u5982\u679c\u60a8\u6709\u4f18\u79c0\u7684\u5de5\u4f5c\u60f3\u8981\u5206\u4eab\uff0c\u6b22\u8fce\u6295\u7a3f\u6216\u8005\u8054\u7cfb\u62a5\u9053\u3002\u6295\u7a3f\u90ae\u7bb1\uff1aliyazhou@jiqizhixin.com\uff1bzhaoyunfeng@jiqizhixin.com<\/section>\n<\/section>\n<\/blockquote>\n<p><strong>\u5317\u4eac\u5927\u5b66\u7684\u5de5\u4f5c\u6765\u81ea\u667a\u80fd\u5b66\u9662\u738b\u7acb\u5a01\u3001\u8d3a\u7b1b\u8001\u5e08\u8bfe\u9898\u7ec4\uff0c\u53d1\u8868\u4e8eICML 2024\u3002\u4f5c\u8005\u5305\u62ec\u5317\u4eac\u5927\u5b66\u56fe\u7075\u73ed\u672c\u79d1\u751f\u6768\u94e0\uff1b\u82cf\u9ece\u4e16\u8054\u90a6\u7406\u5de5\u5b66\u9662\u7855\u58eb\u751fJan Ackermann\uff1b\u5317\u4eac\u5927\u5b66\u667a\u80fd\u5b66\u9662\u535a\u58eb\u751f\u4f55\u9707\u5b87\u3001\u51af\u53e4\u8c6a\u3001\u5f20\u535a\u822a\uff1b\u7ebd\u7ea6\u5927\u5b66\u535a\u58eb\u751f\u51af\u97eb\u799b\uff1b\u5317\u4eac\u667a\u6e90\u7814\u7a76\u9662\u7814\u7a76\u5458\u53f6\u542f\u5a01\uff1b\u6e05\u534e\u5927\u5b66\u7684\u5de5\u4f5c\u6765\u81ea\u4e8e\u660e\u5e74\u5373\u5c06\u5165\u804c\u6e05\u534e\u5927\u5b66\u4ea4\u53c9\u4fe1\u606f\u9662\u505a\u52a9\u7406\u6559\u6388\u3001\u76ee\u524d\u5728\u52a0\u5dde\u5927\u5b66\u4f2f\u514b\u5229\u5206\u6821 Simons Institute \u505a\u535a\u58eb\u540e\u7684\u5415\u51ef\u98ce\u7814\u7a76\u5458\u3002\u4f5c\u8005\u5305\u62ec\u65af\u5766\u798f\u5728\u8bfb\u4e00\u5e74\u7ea7\u535a\u58eb\u6e29\u51ef\u8d8a\uff1b\u6e05\u534e\u5927\u5b66\u59da\u73ed\u672c\u79d1\u751f\u515a\u661f\u5b87\u3002\u00a0<\/strong><\/p>\n<section>\u601d\u7ef4\u94fe\uff08CoT\uff09\u662f\u5927\u6a21\u578b\u4e2d\u6700\u795e\u79d8\u7684\u73b0\u8c61\u4e4b\u4e00\uff0c\u5c24\u5176\u5728\u6570\u5b66\u4efb\u52a1\u4e0a\u663e\u8457\u63d0\u5347\u4e86 Transformer \u7684\u80fd\u529b\u3002\u7136\u800c\uff0c\u601d\u7ef4\u94fe\u7684\u5f15\u5165\u4e5f\u4f7f\u751f\u6210\u5185\u5bb9\u7684\u957f\u5ea6\u589e\u52a0\uff0c\u6d88\u8017\u4e86\u66f4\u591a\u7684\u8ba1\u7b97\u8d44\u6e90\u3002\u8fd9\u4e0d\u7981\u8ba9\u4eba\u597d\u5947\uff1a\u6700\u65b0\u63a8\u51fa\u7684\u9ad8\u6548\u6a21\u578b\uff08\u5982 Mamba\uff09\u662f\u5426\u4e5f\u80fd\u50cf Transformer \u4e00\u6837\u5177\u5907\u5f3a\u5927\u7684\u63a8\u7406\u80fd\u529b\uff1f\u8fd1\u671f\uff0c\u5317\u5927\u548c\u6e05\u534e\u7684\u7814\u7a76\u56e2\u961f\u540c\u65f6\u7ed9\u51fa\u4e86\u660e\u786e\u7684\u5426\u5b9a\u7b54\u6848\uff0c\u63ed\u793a\u4e86 Mamba \u7b49\u9ad8\u6548\u6a21\u578b\u5728\u7ed3\u6784\u4e0a\u7684\u5c40\u9650\u6027\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-fce503ded2835546f3f09f7bde1dd356.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-fce503ded2835546f3f09f7bde1dd356.png\"><\/a><\/p>\n<ul>\n<li>\n<section>\u8bba\u6587 1\uff1aDo Efficient Transformers Really Save Computation? \uff08\u53d1\u8868\u4e8e ICML 2024\uff09<\/section>\n<\/li>\n<li>\n<section>\u8bba\u6587\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/2402.13934<\/section>\n<\/li>\n<\/ul>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-c9fcfb80d2a83de8f8e856073634be10.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-c9fcfb80d2a83de8f8e856073634be10.png\"><\/a><\/p>\n<ul>\n<li>\n<section>\u8bba\u6587 2\uff1aRNNs are not Transformers (Yet): The Key Bottleneck on In-context Retrieval<\/section>\n<\/li>\n<li>\n<section>\u8bba\u6587\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/2402.18510<\/section>\n<\/li>\n<\/ul>\n<section><\/section>\n<section>\u57fa\u4e8e Transformer \u7684\u5927\u8bed\u8a00\u6a21\u578b\u5728\u6587\u672c\u7ffb\u8bd1\u3001\u6587\u672c\u751f\u6210\u7b49\u8bb8\u591a\u9886\u57df\u5c55\u73b0\u4e86\u60ca\u4eba\u7684\u80fd\u529b\u3002\u4e3b\u6d41\u7684\u5927\u8bed\u8a00\u6a21\u578b\u901a\u5e38\u91c7\u7528\u81ea\u56de\u5f52\u8303\u5f0f\u8fdb\u884c\u751f\u6210\uff1a\u7531\u95ee\u9898\u63cf\u8ff0\u3001\u76f8\u5173\u63d0\u793a\u7ec4\u6210\u7684\u8f93\u5165\u5e8f\u5217\uff08prompt\uff09\u4f1a\u88ab\u9996\u5148\u7f16\u7801\u3002\u57fa\u4e8e\u7f16\u7801\u540e\u7684\u4fe1\u606f\uff0c\u5927\u6a21\u578b\u9010\u6b65\u751f\u6210\u540e\u7eed\u7684\u5355\u8bcd\u5e8f\u5217\uff0c\u4ee5\u5f62\u6210\u95ee\u9898\u7684\u7b54\u6848\u3002\u5bf9\u4e8e\u590d\u6742\u7684\u95ee\u9898\uff0c\u5df2\u6709\u7684\u5b9e\u8df5\u548c\u7406\u8bba\u7814\u7a76\u8868\u660e\uff0c\u5229\u7528\u601d\u7ef4\u94fe\u63d0\u793a\uff08CoT\uff09\u53ef\u4ee5\u663e\u8457\u63d0\u5347\u6a21\u578b\u5728\u6570\u5b66\u6216\u63a8\u7406\u65b9\u9762\u7684\u95ee\u9898\u89e3\u51b3\u80fd\u529b\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-a4b3a15d38c7ba1ded60a91f59d62838.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-a4b3a15d38c7ba1ded60a91f59d62838.png\"><\/a><\/p>\n<section>\u5982\u4e0a\u56fe\uff0c\u5f53\u7ed9\u51fa\u7684\u95ee\u9898\u793a\u4f8b\u4e2d\u53ea\u6709\u6700\u7ec8\u7ed3\u679c\uff0c\u800c\u6ca1\u6709\u4e2d\u95f4\u7684\u63a8\u7406\u8fc7\u7a0b\u65f6\uff0c\u5927\u6a21\u578b\u5728\u89c1\u5230\u65b0\u95ee\u9898\u65f6\u5f80\u5f80\u4f1a\u751f\u6210\u9519\u8bef\u7684\u7b54\u6848\uff1b\u4f46\u5982\u679c\u7ed9\u51fa\u7684\u95ee\u9898\u793a\u4f8b\u4e2d\u5305\u542b\u4e86\u5b8c\u6574\u7684\u601d\u7ef4\u94fe\uff08\u4e2d\u95f4\u63a8\u5bfc\u6b65\u9aa4\uff09\uff0c\u5927\u6a21\u578b\u5728\u65b0\u95ee\u9898\u4e0a\u4e5f\u5f80\u5f80\u4f1a\u8f93\u51fa\u5b8c\u6574\u7684\u601d\u7ef4\u94fe\u5e76\u5f97\u5230\u6b63\u786e\u7b54\u6848\u3002\u7136\u800c\uff0c\u601d\u7ef4\u94fe\u63d0\u793a\u4f9d\u8d56\u4e8e\u5927\u91cf\u4e2d\u95f4\u6b65\u9aa4\u7684\u8f93\u51fa\uff0c\u4f7f\u7528 Transformer \u67b6\u6784\u4f1a\u4ea7\u751f\u5927\u91cf\u7684\u8ba1\u7b97\u5f00\u9500\u3002\u56e0\u6b64\uff0c\u4e00\u4e2a\u81ea\u7136\u7684\u95ee\u9898\u4fbf\u662f\uff1a\u80fd\u5426\u4f7f\u7528\u66f4\u5c11\u7684\u8ba1\u7b97\u91cf\uff0c\u5b9e\u73b0\u601d\u7ef4\u94fe\u63a8\u7406\uff1f\u8bb8\u591a\u67b6\u6784\u81f4\u529b\u4e8e\u964d\u4f4e\u6ce8\u610f\u529b\u673a\u5236\u5e26\u6765\u7684\u8ba1\u7b97\u590d\u6742\u5ea6\uff0c\u80fd\u5426\u5728\u8fd9\u4e9b\u67b6\u6784\u4e0a\u4f7f\u7528 CoT \u63d0\u5347\u6027\u80fd\u5e76\u5b9e\u73b0\u63a8\u7406\u7684\u52a0\u901f\uff1f<\/section>\n<section><\/section>\n<section>\u8fd1\u671f\uff0c\u6765\u81ea\u5317\u5927\u548c\u6e05\u534e\u7684\u7814\u7a76\u56e2\u961f\u4ece\u7406\u8bba\u89d2\u5ea6\u5bf9\u4e0a\u8ff0\u95ee\u9898\u8fdb\u884c\u4e86\u6df1\u5165\u63a2\u8ba8\u3002\u7ed3\u679c\u4ee4\u4eba\u60ca\u8bb6\uff1a<strong>\u4e24\u4e2a\u56e2\u961f\u4e00\u81f4\u8bc1\u5b9e\u5305\u62ec Sparse Transformer\u3001Linear Transformer\u3001Mamba \u5728\u5185\u7684\u8bb8\u591a\u67b6\u6784\uff0c\u5373\u4f7f\u5728\u8fd9\u4e9b\u6a21\u578b\u4e0a\u5e94\u7528\u601d\u7ef4\u94fe\uff0c\u5176\u7406\u8bba\u4e0a\u7684\u80fd\u529b\u4e0a\u9650\u4ecd\u65e0\u6cd5\u89e3\u51b3\u591a\u79cd\u5b9e\u9645\u63a8\u7406\u95ee\u9898\uff0c\u5e76\u4e0e\u6807\u51c6 Transformer \u6709\u672c\u8d28\u5dee\u8ddd\u3002\u8fd9\u4e9b\u7406\u8bba\u7ed3\u679c\u4e3a\u9ad8\u6548\u7ed3\u6784\u7684\u5b9e\u7528\u4ef7\u503c\u8499\u4e0a\u4e86\u4e00\u5c42\u9634\u5f71\u3002<\/strong><\/section>\n<section><\/section>\n<section><strong>Transformer + CoT \u4f9d\u7136\u662f\u6700\u4f73\u9009\u9879<\/strong><\/section>\n<section><\/section>\n<section>\u5317\u5927\u7684\u7814\u7a76\u56e2\u961f\u5c06\u63a8\u7406\u4efb\u52a1\u901a\u7528\u5730\u5efa\u6a21\u4e3a\u52a8\u6001\u89c4\u5212\u3002\u5728\u63a8\u7406\u8fc7\u7a0b\u4e2d\uff0c\u6a21\u578b\u9700\u8981\u6309\u7167\u5408\u6cd5\u7684\u62d3\u6251\u6392\u5e8f\u9010\u6b65\u8f93\u51fa\u5b50\u95ee\u9898\u7684\u7ed3\u679c\uff08\u5373\u601d\u7ef4\u94fe\uff09\uff0c\u4ee5\u6700\u7ec8\u5f97\u51fa\u539f\u95ee\u9898\u7684\u89e3\u3002\u4ed6\u4eec\u4ece\u7406\u8bba\u89d2\u5ea6\u8bc1\u660e\uff0c\u82e5\u5e0c\u671b Sparse Transformers\u3001Linear Transformers \u7b49\u6a21\u578b\u901a\u8fc7\u601d\u7ef4\u94fe\u6765\u89e3\u51b3\u52a8\u6001\u89c4\u5212\u95ee\u9898\uff0c\u6a21\u578b\u7684\u5bbd\u5ea6\u5fc5\u987b\u589e\u52a0\uff0c\u8fdb\u800c\u5bfc\u81f4\u65f6\u95f4\u590d\u6742\u5ea6\u6070\u597d\u8fbe\u5230<strong>\u5e73\u65b9\u7ea7\u522b<\/strong>\u3002\u8fd9\u4e00\u65f6\u95f4\u590d\u6742\u5ea6\u4e0e\u6807\u51c6 Transformers \u5728\u76f8\u540c\u957f\u5ea6\u7684\u63a8\u7406\u4efb\u52a1\u4e2d\u6240\u9700\u7684\u590d\u6742\u5ea6\u4e00\u81f4\uff0c\u8868\u660e<strong>\u8fd9\u4e9b\u6240\u8c13\u7684\u9ad8\u6548\u7ed3\u6784\u5728\u4e00\u822c\u63a8\u7406\u95ee\u9898\u4e0a\u5e76\u4e0d\u5177\u5907\u8ba1\u7b97\u4f18\u52bf\u3002<\/strong><\/section>\n<section><\/section>\n<section>\u6b8a\u9014\u540c\u5f52\uff0c\u6e05\u534e\u7684\u7814\u7a76\u56e2\u961f\u8003\u8651\u4e86 RNN \u6a21\u578b\u548c Transformer \u6a21\u578b\u5728\u68c0\u7d22\u3001\u5173\u8054\u56de\u5fc6\u3001\u8ba1\u6570\uff0c\u4ee5\u53ca\u5224\u65ad\u4e00\u5f20\u56fe\u662f\u5426\u4e3a\u6811\u7b49\u57fa\u672c\u95ee\u9898\u4e0a\u7684\u8868\u8fbe\u80fd\u529b\u5dee\u5f02\u3002\u4ed6\u4eec\u4ece\u7406\u8bba\u89d2\u5ea6\u8bc1\u660e\u4e86\uff0c\u5bf9\u4e8e\u95ee\u9898\u89c4\u6a21 n\uff0c\u4efb\u610f o (n) \u5927\u5c0f\u7684 RNN \u6a21\u578b\u5747\u65e0\u6cd5\u5b8c\u6210\u4e0a\u8ff0\u4efb\u52a1\uff0c\u5373\u4fbf\u4f7f\u7528\u4efb\u610f\u957f\u7684 CoT\u3002\u4f46\u4e00\u4e2a\u56fa\u5b9a\u5927\u5c0f\u7684 Transformer \u53ef\u4ee5\u4e0d\u4f7f\u7528 CoT \u89e3\u51b3\u68c0\u7d22\u3001\u5173\u8054\u56de\u5fc6\u3001\u8ba1\u6570\u7b49\u95ee\u9898\uff0c\u5e76\u4f7f\u7528 O (n) \u957f\u5ea6\u7684 CoT \u6b63\u786e\u5224\u65ad\u4e00\u5f20\u56fe\u662f\u5426\u4e3a\u6811\u3002\u8fd9\u4e9b\u7ed3\u679c\u8868\u660e<strong>\u7c7b RNN \u7684\u7ed3\u6784\u4f7f\u7528\u601d\u7ef4\u94fe\u83b7\u5f97\u7684\u80fd\u529b\uff0c\u8ddd\u79bb Transformer \u4ecd\u6709\u5de8\u5927\u5dee\u8ddd<\/strong>\u3002<\/section>\n<section><\/section>\n<section>\u4e24\u4e2a\u7814\u7a76\u5f97\u51fa\u76f8\u4f3c\u7ed3\u8bba\u7684\u80cc\u540e\uff0c\u63ed\u793a\u4e86\u76f8\u4f3c\u7684\u672c\u8d28\u3002\u8fd9\u4e9b\u6a21\u578b\u67b6\u6784\u4e4b\u6240\u4ee5\u5177\u6709\u5185\u5b58\u9ad8\u6548\u6027\uff0c\u662f\u56e0\u4e3a\u5b83\u4eec\u80fd\u591f\u5728\u8f83\u5c11\u7684\u7a7a\u95f4\u4e0b\u5b8c\u5168\u786e\u5b9a\u8f93\u51fa\u5e8f\u5217\u3002\u7136\u800c\uff0c\u8f93\u51fa\u5e8f\u5217\u7684\u53ef\u80fd\u6027\u6307\u6570\u7ea7\u589e\u957f\uff0c\u8fd9\u4f7f\u5f97\u8fd9\u4e9b\u67b6\u6784\u65e0\u6cd5\u4ee5\u8f83\u5c0f\u7684\u6a21\u578b\u5c3a\u5bf8\u6b63\u786e\u751f\u6210\u6240\u6709\u53ef\u80fd\u7684\u8f93\u51fa\u3002\u6362\u53e5\u8bdd\u8bf4\uff0c\u6a21\u578b\u7684\u89c4\u6a21\u5fc5\u987b\u968f\u7740\u95ee\u9898\u89c4\u6a21\u7684\u589e\u52a0\u800c\u6269\u5c55\u3002\u8fd9\u4e00\u6838\u5fc3\u89c2\u5bdf\u8868\u660e\uff0c<strong>\u5305\u62ec Mamba \u5728\u5185\u7684\u5177\u6709\u5faa\u73af\uff08recurrent\uff09\u6027\u8d28\u7684\u7f51\u7edc\u67b6\u6784\uff0c\u90fd\u53d7\u5230\u4e86\u76f8\u540c\u7684\u9650\u5236\u3002<\/strong><\/section>\n<section><\/section>\n<section><strong>\u5982\u4f55\u63d0\u5347\u5176\u5b83\u6a21\u578b\u4f7f\u7528\u601d\u7ef4\u94fe\u7684\u80fd\u529b\uff1f<\/strong><\/section>\n<section><\/section>\n<section>\u5728\u5f97\u5230\u8d1f\u9762\u7ed3\u8bba\u7684\u540c\u65f6\uff0c\u4e24\u4e2a\u56e2\u961f\u8003\u8651\u4e0d\u540c\u89d2\u5ea6\u8bbe\u8ba1\u89e3\u51b3\u65b9\u6848\u3002<\/section>\n<section><\/section>\n<section>\u5317\u5927\u7814\u7a76\u56e2\u961f\u4ece\u63a8\u7406\u4efb\u52a1\u7684\u5c40\u90e8\u6027\u5165\u624b\uff0c\u5373\u5f53\u524d\u8f93\u51fa\u6240\u9700\u8981\u5411\u524d\u4f9d\u8d56\u7684\u6700\u8fdc\u8f93\u51fa\u7684\u8ddd\u79bb\u3002\u5c40\u90e8\u6027\u53cd\u6620\u4e86\u63a8\u7406\u6240\u9700\u8981\u7684\u957f\u7a0b\u8bb0\u5fc6\u96be\u5ea6\u3002\u4ed6\u4eec\u4ece\u7406\u8bba\u89d2\u5ea6\u8bc1\u660e\u4e86\uff0c\u5f53\u63a8\u7406\u4efb\u52a1\u6709\u8f83\u597d\u7684\u5c40\u90e8\u6027\u65f6\uff0c\u8bb8\u591a\u9ad8\u6548\u6a21\u578b\u80fd\u591f\u4ee5\u7406\u8bba\u66f4\u4f18\u7684\u63a8\u7406\u901f\u5ea6\u5b8c\u7f8e\u89e3\u51b3\u95ee\u9898\u3002<\/section>\n<section><\/section>\n<section>\u6765\u81ea\u6e05\u534e\u7684\u7814\u7a76\u56e2\u961f\u5219\u4ece\u5f15\u5165\u4e0a\u4e0b\u6587\u68c0\u7d22\u5668\u7684\u89d2\u5ea6\u5165\u624b\u3002\u4ed6\u4eec\u4ece\u7406\u8bba\u89d2\u5ea6\u8bc1\u660e\u4e86\uff0c\u4f7f\u7528\u663e\u5f0f\u4e0a\u4e0b\u6587\u68c0\u7d22\u5668\u6216\u4f7f\u7528\u4e00\u5c42 Transformer \u4f5c\u4e3a\u9690\u5f0f\u4e0a\u4e0b\u6587\u68c0\u7d22\u5668\uff0c\u5747\u53ef\u4ee5\u5927\u5927\u589e\u5f3a RNN \u4f7f\u7528\u601d\u7ef4\u94fe\u540e\u83b7\u5f97\u7684\u80fd\u529b\u3002<\/section>\n<section><\/section>\n<section>\u7814\u7a76\u56e2\u961f\u8fd8\u8bbe\u8ba1\u4e86\u5927\u91cf\u7684\u5b9e\u9a8c\u9a8c\u8bc1\u7406\u8bba\u7ed3\u679c\u3002\u6765\u81ea\u5317\u5927\u7684\u7814\u7a76\u56e2\u961f\u5728\u591a\u79cd Transformer \u67b6\u6784\u4e0a\u4f7f\u7528 CoT \u6570\u636e\u8fdb\u884c\u8bad\u7ec3\uff0c\u8868\u660e\u6807\u51c6 Transformer \u67b6\u6784\u4e8b\u5b9e\u4e0a\u4f7f\u7528\u4e86\u6700\u5c0f\u7684\u8ba1\u7b97\u91cf\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-a156c678b3576a34bd7b2adbeb6c28e2.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-a156c678b3576a34bd7b2adbeb6c28e2.png\"><\/a><\/p>\n<section>\u540c\u65f6\uff0c\u4ed6\u4eec\u7684\u5b9e\u9a8c\u7ed3\u679c\u4e5f\u8868\u660e\u4e86\u826f\u597d\u7684\u63a8\u7406\u5c40\u90e8\u6027\uff08\u4e0b\u56fe\u53f3\u5217\uff09\u80fd\u591f\u589e\u5f3a\u591a\u79cd Efficient Transformers \u7684\u601d\u7ef4\u94fe\u63a8\u7406\u80fd\u529b\uff0c\u8868\u73b0\u5728\u4f7f\u7528\u76f8\u540c\u5c3a\u5bf8\u7684\u6a21\u578b\u65f6\u80fd\u591f\u89e3\u51b3\u89c4\u6a21\u66f4\u5927\u7684\u63a8\u7406\u4efb\u52a1\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-80671f3d4069a2ff23480086ee426b4f.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-80671f3d4069a2ff23480086ee426b4f.png\"><\/a><\/p>\n<section>\u6765\u81ea\u6e05\u534e\u7684\u7814\u7a76\u56e2\u961f\u5219\u5728 Mamba \u548c Transformer \u4e0a\u4f7f\u7528\u5224\u5b9a\u56fe\u662f\u5426\u4e3a\u6811\u7684\u4efb\u52a1\u8fdb\u884c\u8bad\u7ec3\uff0c\u8868\u660e Transformer \u7ed3\u6784\u5728\u6b64\u95ee\u9898\u4e0a\u76f8\u6bd4\u4e8e Mamba \u6240\u5177\u6709\u7684\u663e\u8457\u4f18\u52bf\uff0c\u4ee5\u53ca\u5728 RNN \u4e0a\u4f7f\u7528\u4e0a\u4e0b\u6587\u68c0\u7d22\u5bf9\u6027\u80fd\u7684\u5f71\u54cd\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-699896dc0de834e7f75e0495c1bf1e00.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-699896dc0de834e7f75e0495c1bf1e00.png\"><\/a><\/p>\n<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-10-21\" target=\"_blank\">\u8fd8\u662f\u539f\u88c5Transformer\u597d\uff01\u5317\u5927\u6e05\u534e\u56e2\u961f\u540c\u65f6\u63ed\u793aMamba\u7b49\u63a8\u7406\u77ed\u677f<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:\u8fd8\u662f\u539f\u88c5Transfo [&hellip;]<\/p>\n","protected":false},"author":3,"featured_media":0,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[27],"tags":[68],"class_list":["post-36227","post","type-post","status-publish","format-standard","hentry","category-news","tag-68"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.9 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>\u8fd8\u662f\u539f\u88c5Transformer\u597d\uff01\u5317\u5927\u6e05\u534e\u56e2\u961f\u540c\u65f6\u63ed\u793aMamba\u7b49\u63a8\u7406\u77ed\u677f - \u4e00\u8d77AI\u6280\u672f<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/17aitech.com\/?p=36227\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=36227#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=36227\"},\"author\":{\"name\":\"AI\u5c0f\u52a9\u624b\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\"},\"headline\":\"\u8fd8\u662f\u539f\u88c5Transformer\u597d\uff01\u5317\u5927\u6e05\u534e\u56e2\u961f\u540c\u65f6\u63ed\u793aMamba\u7b49\u63a8\u7406\u77ed\u677f\",\"datePublished\":\"2025-01-07T16:00:22+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=36227\"},\"wordCount\":96,\"commentCount\":0,\"image\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=36227#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/10\\\/frc-43abb2d37958289a28016caa2c553119.png\",\"keywords\":[\"\u5927\u6a21\u578b\"],\"articleSection\":[\"\u884c\u4e1a\u8d44\u8baf\"],\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/17aitech.com\\\/?p=36227#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=36227\",\"url\":\"https:\\\/\\\/17aitech.com\\\/?p=36227\",\"name\":\"\u8fd8\u662f\u539f\u88c5Transformer\u597d\uff01\u5317\u5927\u6e05\u534e\u56e2\u961f\u540c\u65f6\u63ed\u793aMamba\u7b49\u63a8\u7406\u77ed\u677f - \u4e00\u8d77AI\u6280\u672f\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=36227#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=36227#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/10\\\/frc-43abb2d37958289a28016caa2c553119.png\",\"datePublished\":\"2025-01-07T16:00:22+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\"},\"breadcrumb\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=36227#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/17aitech.com\\\/?p=36227\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=36227#primaryimage\",\"url\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/10\\\/frc-43abb2d37958289a28016caa2c553119.png\",\"contentUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/10\\\/frc-43abb2d37958289a28016caa2c553119.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=36227#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\\\/\\\/17aitech.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u8fd8\u662f\u539f\u88c5Transformer\u597d\uff01\u5317\u5927\u6e05\u534e\u56e2\u961f\u540c\u65f6\u63ed\u793aMamba\u7b49\u63a8\u7406\u77ed\u677f\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/#website\",\"url\":\"https:\\\/\\\/17aitech.com\\\/\",\"name\":\"\u4e00\u8d77AI\u6280\u672f\",\"description\":\"\u8ba9AI\u77e5\u8bc6\u89e6\u624b\u53ef\u53ca\",\"alternateName\":\"\u4e00\u8d77AI\u6280\u672f\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/17aitech.com\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\",\"name\":\"AI\u5c0f\u52a9\u624b\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/04\\\/robot_3.png\",\"url\":\"\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/04\\\/robot_3.png\",\"contentUrl\":\"\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/04\\\/robot_3.png\",\"caption\":\"AI\u5c0f\u52a9\u624b\"},\"description\":\"\u8fd9\u4e2a\u4eba\u5f88\u61d2\uff0c\u4ec0\u4e48\u90fd\u6ca1\u6709\u7559\u4e0b\uff5e\",\"url\":\"https:\\\/\\\/17aitech.com\\\/?page_id=33738&user=3\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u8fd8\u662f\u539f\u88c5Transformer\u597d\uff01\u5317\u5927\u6e05\u534e\u56e2\u961f\u540c\u65f6\u63ed\u793aMamba\u7b49\u63a8\u7406\u77ed\u677f - \u4e00\u8d77AI\u6280\u672f","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/17aitech.com\/?p=36227","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/17aitech.com\/?p=36227#article","isPartOf":{"@id":"https:\/\/17aitech.com\/?p=36227"},"author":{"name":"AI\u5c0f\u52a9\u624b","@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"headline":"\u8fd8\u662f\u539f\u88c5Transformer\u597d\uff01\u5317\u5927\u6e05\u534e\u56e2\u961f\u540c\u65f6\u63ed\u793aMamba\u7b49\u63a8\u7406\u77ed\u677f","datePublished":"2025-01-07T16:00:22+00:00","mainEntityOfPage":{"@id":"https:\/\/17aitech.com\/?p=36227"},"wordCount":96,"commentCount":0,"image":{"@id":"https:\/\/17aitech.com\/?p=36227#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-43abb2d37958289a28016caa2c553119.png","keywords":["\u5927\u6a21\u578b"],"articleSection":["\u884c\u4e1a\u8d44\u8baf"],"inLanguage":"zh-Hans","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/17aitech.com\/?p=36227#respond"]}]},{"@type":"WebPage","@id":"https:\/\/17aitech.com\/?p=36227","url":"https:\/\/17aitech.com\/?p=36227","name":"\u8fd8\u662f\u539f\u88c5Transformer\u597d\uff01\u5317\u5927\u6e05\u534e\u56e2\u961f\u540c\u65f6\u63ed\u793aMamba\u7b49\u63a8\u7406\u77ed\u677f - \u4e00\u8d77AI\u6280\u672f","isPartOf":{"@id":"https:\/\/17aitech.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/17aitech.com\/?p=36227#primaryimage"},"image":{"@id":"https:\/\/17aitech.com\/?p=36227#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-43abb2d37958289a28016caa2c553119.png","datePublished":"2025-01-07T16:00:22+00:00","author":{"@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"breadcrumb":{"@id":"https:\/\/17aitech.com\/?p=36227#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/17aitech.com\/?p=36227"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/?p=36227#primaryimage","url":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-43abb2d37958289a28016caa2c553119.png","contentUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-43abb2d37958289a28016caa2c553119.png"},{"@type":"BreadcrumbList","@id":"https:\/\/17aitech.com\/?p=36227#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/17aitech.com\/"},{"@type":"ListItem","position":2,"name":"\u8fd8\u662f\u539f\u88c5Transformer\u597d\uff01\u5317\u5927\u6e05\u534e\u56e2\u961f\u540c\u65f6\u63ed\u793aMamba\u7b49\u63a8\u7406\u77ed\u677f"}]},{"@type":"WebSite","@id":"https:\/\/17aitech.com\/#website","url":"https:\/\/17aitech.com\/","name":"\u4e00\u8d77AI\u6280\u672f","description":"\u8ba9AI\u77e5\u8bc6\u89e6\u624b\u53ef\u53ca","alternateName":"\u4e00\u8d77AI\u6280\u672f","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/17aitech.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d","name":"AI\u5c0f\u52a9\u624b","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","url":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","contentUrl":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","caption":"AI\u5c0f\u52a9\u624b"},"description":"\u8fd9\u4e2a\u4eba\u5f88\u61d2\uff0c\u4ec0\u4e48\u90fd\u6ca1\u6709\u7559\u4e0b\uff5e","url":"https:\/\/17aitech.com\/?page_id=33738&user=3"}]}},"_links":{"self":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/36227","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=36227"}],"version-history":[{"count":0,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/36227\/revisions"}],"wp:attachment":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=36227"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=36227"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=36227"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}