{"id":9486,"date":"2025-11-11T17:04:00","date_gmt":"2025-11-11T09:04:00","guid":{"rendered":"\/?p=9486"},"modified":"2025-11-11T17:21:45","modified_gmt":"2025-11-11T09:21:45","slug":"gtc25-%e5%ad%97%e8%8a%82%e5%8f%82%e4%bc%9a%e6%96%87%e7%ab%a0-text2sql-%e6%96%b0%e8%af%84%e4%bc%b0%e6%8c%87%e6%a0%87-funcevalgmn","status":"publish","type":"post","link":"\/?p=9486","title":{"rendered":"GTC25 | \u5b57\u8282\u53c2\u4f1a\u6587\u7ae0 Text2SQL \u65b0\u8bc4\u4f30\u6307\u6807-FuncEvalGMN"},"content":{"rendered":"<p>\u672c\u6587\u5c06\u4ecb\u7ecd GTC 2025 session \u4e4b\u4e00\uff0c\u7531\u5b57\u8282\u56e2\u961f\u63d0\u51fa\u7684\u57fa\u4e8e\u56fe\u7684 Text2SQL \u7684\u65b0\u8bc4\u4f30\u6307\u6807\uff0c\u6807\u9898\u4e3a\uff1aToward Database-Free Text-to-SQL Evaluation: A Graph-Based Metric for Functional Correctness [P72633]<img src=\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/640-20251111170312619.png\" alt=\"\u56fe\u7247\" \/><\/p>\n<p>\u662f\u5b57\u8282\u56e2\u961f\u63d0\u51fa\u7684<strong>\u57fa\u4e8e\u56fe<\/strong>\u7684\u65b0\u8bc4\u4f30\u6307\u6807-FuncEvalGMN\uff0c\u4e3b\u8981\u662f\u4e3a\u4e86\u514b\u670d\u6267\u884c\u51c6\u786e\u6027\uff08Execution Accuracy\uff09\u548c\u7cbe\u786e\u96c6\u5408\u5339\u914d\uff08Exact Set Match\uff09\u7684\u5c40\u9650\u6027\u3002<\/p>\n<p>FuncEvalGMN \u6709\u4e24\u4e2a\u4f18\u52bf\uff1a\u4e00\u662f<strong>\u53ea\u9700\u8981\u6570\u636e\u5e93 schema<\/strong> \u5c31\u80fd\u63a8\u5bfc\u51fa\u903b\u8f91\u6267\u884c\u8ba1\u5212\uff0c\u65e0\u9700\u5927\u91cf\u51c6\u5907\u6d4b\u8bd5\u6570\u636e\u5e93\uff1b\u4e8c\u662f\u5728<strong>\u672a\u89c1\u6570\u636e\u96c6\u4e0a\u6709\u5f88\u5f3a\u7684\u6cdb\u5316\u80fd\u529b<\/strong>\uff0c\u51f8\u663e\u4e86\u5176\u51c6\u786e\u6027\u548c\u7a33\u5065\u6027\u3002<\/p>\n<p>\u4e0b\u9762\u5c06\u5206\u4eab\u4ea7\u54c1\u7ec6\u8282\uff0c\u60a8\u53ef\u4ee5\u901a\u8fc7\u4ee5\u4e0b\u94fe\u63a5\u627e\u5230\u672c\u6587\u7684\u6240\u6709\u53c2\u8003\u5185\u5bb9<\/p>\n<blockquote>\n<p><strong>\u6a21\u578b\u6743\u91cd\uff1a<\/strong><a href=\"https:\/\/drive.google.com\/drive\/folders\/1KnVtwlDuIExoEY3Bq7ayKHhldhCtIydJ\">https:\/\/drive.google.com\/drive\/folders\/1KnVtwlDuIExoEY3Bq7ayKHhldhCtIydJ<\/a><\/p>\n<p><strong>GTC session \u94fe\u63a5\uff1a<\/strong><a href=\"https:\/\/www.nvidia.cn\/gtc-global\/session-catalog\/?tab.catalogallsessionstab=16566177511100015Kus&amp;search=sql#\/session\/1727580017854001n7SC\">https:\/\/www.nvidia.cn\/gtc-global\/session-catalog\/?tab.catalogallsessionstab=16566177511100015Kus&search=sql#\/session\/1727580017854001n7SC<\/a><\/p>\n<p><strong>\u5168\u6587pdf\uff1a<\/strong><a href=\"https:\/\/aclanthology.org\/2025.coling-main.308.pdf\">https:\/\/aclanthology.org\/2025.coling-main.308.pdf<\/a><\/p>\n<p><strong>\u4ee3\u7801\uff1a<\/strong><a href=\"https:\/\/github.com\/Leon0-0\/FuncEvalGMN\">https:\/\/github.com\/Leon0-0\/FuncEvalGMN<\/a><\/p>\n<\/blockquote>\n<hr \/>\n<p>GTC session \u4ecb\u7ecd\uff1a<\/p>\n<p><img src=\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/640-20251111170313194.png\" alt=\"\u56fe\u7247\" \/><\/p>\n<p>\u67b6\u6784\u56fe\uff1a<\/p>\n<p>\u9996\u5148\uff0c\u5c06 SQL \u67e5\u8be2\u5bf9\u89e3\u6790\u4e3a\u56fe\u5e76\u5d4c\u5165\u5176\u8282\u70b9\u3002\u7136\u540e\uff0c\u4f7f\u7528\u56fe\u5bf9\u6bd4\u5b66\u4e60\u6765\u8bad\u7ec3\u56fe\u5339\u914d\u7f51\u7edc\u3002\u6700\u540e\uff0c\u57fa\u4e8e\u56fe\u8868\u793a\u7684\u76f8\u4f3c\u6027\u6765\u8bc4\u4f30\u751f\u6210\u7684 SQL \u7684\u6b63\u786e\u6027\u3002<\/p>\n<p><img src=\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/640-20251111170313436.png\" alt=\"\u56fe\u7247\" \/><\/p>\n<p>\u6838\u5fc3\u662f\u5229\u7528<strong>\u5173\u7cfb\u8fd0\u7b97\u7b26\u6811\uff08ROT\uff09<\/strong>\u4ece SQL \u67e5\u8be2\u7684<strong>\u903b\u8f91\u6267\u884c\u8ba1\u5212<\/strong>\u4e2d\u63d0\u53d6\u4e30\u5bcc\u7684\u8bed\u4e49\u4fe1\u606f\u5e76\u5d4c\u5165\u5230\u56fe\u4e2d\uff0c\u901a\u8fc7\u56fe\u5bf9\u6bd4\u5b66\u4e60\u8bad\u7ec3<strong>\u56fe\u795e\u7ecf\u7f51\u7edc\uff08GNN\uff09<\/strong>\u5bf9 SQL \u67e5\u8be2\u5bf9\u8fdb\u884c\u56fe\u5339\u914d\u3002<\/p>\n<p>\u4e0b\u56fe\u793a\u610f\uff0c\u8fdb\u884c position embedding\uff1a\u901a\u8fc7\u8fde\u63a5\u56fe\u5bf9\u4e2d\u7684\u5173\u7cfb\u8282\u70b9\uff08Rels\uff09\u548c\u79cd\u5b50\u8282\u70b9\uff08Seed Nodes\uff09\u6765\u8ba1\u7b97\u56fe\u5bf9\u7684\u4f4d\u7f6e\u7f16\u7801<\/p>\n<p><img src=\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/640-20251111170313661.png\" alt=\"\u56fe\u7247\" \/><\/p>\n<p>\u4e0b\u56fe\u793a\u610f\u8fdb\u884c graph embeding\uff1a\u901a\u8fc7\u6574\u5408\u5185\u90e8\u56fe\u6d88\u606f\u4f20\u9012\uff08Inner Graph Message Passing\uff09\u548c\u4ea4\u53c9\u6ce8\u610f\u529b\uff08Cross Attention\uff09\uff0c\u6700\u540e\u8fdb\u884c\u5411\u91cf\u7a7a\u95f4\u7684\u76f8\u4f3c\u5ea6\u8ba1\u7b97\u6765\u5f97\u5230\u7ed3\u679c<\/p>\n<p><img src=\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/640-20251111170316015.png\" alt=\"\u56fe\u7247\" \/><\/p>\n<p>\u6548\u679c\u5bf9\u6bd4\uff1a<\/p>\n<p>FuncEvalGMN \u548c Test Suite \uff08Ruiqi Zhong, Tao Yu, and Dan Klein. 2020. Semantic evaluation for text-to-sql with distilled <strong>test suites<\/strong>. arXiv preprint arXiv:2010.02840.\uff09\u5728\u4e0d\u540c\u6570\u636e\u96c6\u4e0a\u9488\u5bf9\u8fde\u63a5\u67e5\u8be2\u3001\u5d4c\u5957\u67e5\u8be2\u548c\u5206\u7ec4\u67e5\u8be2\u8fdb\u884c\u6027\u80fd\u6bd4\u8f83\uff0c\u4f7f\u7528\u4e86 BIRD-pair dev\u3001Spider-pair dev \u548c Spider-DK-pair dev \u6570\u636e\u96c6\u7684\u7ec4\u5408\u3002\u53ef\u4ee5\u89c2\u5bdf\u5230\uff0cFuncEvalGMN \u65b9\u6cd5\u7684 AUC \u5206\u6570\u59cb\u7ec8\u9ad8\u4e8e\u6d4b\u8bd5\u5957\u4ef6\u3002<\/p>\n<p><img src=\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/640-20251111170318517.png\" alt=\"\u56fe\u7247\" \/><\/p>\n<p><img src=\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/640-20251111170319302.png\" alt=\"\u56fe\u7247\" \/><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u672c\u6587\u5c06\u4ecb\u7ecd GTC 2025 session \u4e4b\u4e00\uff0c\u7531\u5b57\u8282\u56e2\u961f\u63d0\u51fa\u7684\u57fa\u4e8e\u56fe\u7684 Text2SQL \u7684\u65b0\u8bc4\u4f30\u6307\u6807\uff0c\u6807\u9898\u4e3a\uff1aToward Da   \u2026 &#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ngg_post_thumbnail":0},"categories":[288],"tags":[],"_links":{"self":[{"href":"\/index.php?rest_route=\/wp\/v2\/posts\/9486"}],"collection":[{"href":"\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=9486"}],"version-history":[{"count":1,"href":"\/index.php?rest_route=\/wp\/v2\/posts\/9486\/revisions"}],"predecessor-version":[{"id":9487,"href":"\/index.php?rest_route=\/wp\/v2\/posts\/9486\/revisions\/9487"}],"wp:attachment":[{"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=9486"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=9486"},{"taxonomy":"post_tag","embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=9486"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}