
{"id":13266,"date":"2026-10-13T01:26:19","date_gmt":"2026-10-12T17:26:19","guid":{"rendered":"https:\/\/infernews.com\/blog\/programgym\/"},"modified":"2026-10-13T01:26:19","modified_gmt":"2026-10-12T17:26:19","slug":"programgym","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/programgym\/","title":{"rendered":"ProgramGym\uff1a\u7a0b\u5f0f\u91cd\u88fd\u4ee3\u7406\u8a13\u7df4"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/10\/ProgramGym.jpg\" alt=\"ProgramGym logo\"><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u8981\u4ee4 coding agent \u91cd\u5efa\u4e00\u500b\u672a\u516c\u958b\u5be6\u4f5c\u7684\u7a0b\u5f0f\uff0c\u55ae\u9760\u898f\u683c\u4e26\u4e0d\u8db3\u5920\uff0c\u9084\u8981\u63a2\u7d22\u53c3\u8003\u57f7\u884c\u6a94\u3001\u63a8\u65b7\u884c\u70ba\u3001\u8a2d\u8a08\u67b6\u69cb\uff0c\u518d\u53cd\u8986\u4fee\u6b63\u3002ProgramGym \u5c6c\u65bc\u9762\u5411\u7a0b\u5f0f\u91cd\u88fd\u4ee3\u7406\u7684\u8a13\u7df4\u74b0\u5883\u8207\u5de5\u4f5c\u6d41\u7a0b\u9805\u76ee\uff0c\u900f\u904e Program-Env \u548c Program-Agent \u8655\u7406\u6e2c\u8a66\u74b0\u5883\u96e3\u4ee5\u5927\u91cf\u5efa\u7acb\u3001\u9577\u6642\u9593\u958b\u767c\u5bb9\u6613\u5931\u53bb\u8108\u7d61\u5169\u500b\u554f\u984c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Program-Env \u6703\u7531\u771f\u5be6 repository \u5efa\u7acb\u53c3\u8003 executable \u548c runtime sandbox\uff0c\u518d\u4ee5\u9a57\u8b49\u56de\u994b\u9010\u6b65\u7522\u751f behavioral tests\uff1bProgram-Agent \u5247\u628a\u5206\u6790\u3001\u5be6\u4f5c\u548c review \u5206\u958b\u8655\u7406\uff0c\u5229\u7528\u6301\u7e8c\u4fdd\u5b58\u7684\u4ea4\u63a5\u7d00\u9304\u8207 revision checkpoints\uff0c\u8b93\u4ee3\u7406\u53ef\u4ee5\u5ef6\u7e8c\u591a\u8f2a\u5de5\u4f5c\uff0c\u800c\u4e0d\u662f\u6bcf\u6b21\u91cd\u65b0\u7406\u89e3\u6574\u500b\u7a0b\u5f0f\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u9805\u76ee\u76ee\u524d\u63d0\u4f9b ProgramGym-1648 dataset\uff0c\u6db5\u84cb Go\u3001C++\u3001Rust \u548c C\uff0c\u5171 1,648 \u500b\u53ef\u9a57\u8b49\u4efb\u52d9\uff0c\u5176\u4e2d Go \u6709 818 \u500b\u3002\u8cc7\u6599\u96c6\u9069\u5408\u7528\u4f86\u8a13\u7df4\u6216\u6bd4\u8f03 coding agents\uff0c\u5c24\u5176\u662f\u9700\u8981\u9ed1\u76d2\u63a2\u7d22\u3001\u8de8\u591a\u500b\u4fee\u8a02\u968e\u6bb5\uff0c\u4ee5\u53ca\u91cd\u8996\u884c\u70ba\u76f8\u5bb9\u6027\u7684\u5de5\u4f5c\u60c5\u5883\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u2022 Program-Env \u8ca0\u8cac\u5efa\u7acb executable\u3001sandbox \u548c\u884c\u70ba\u6e2c\u8a66<br \/>\n\u2022 Program-Agent \u5206\u62c6\u5206\u6790\u3001\u5be6\u4f5c\u8207 review \u6d41\u7a0b<br \/>\n\u2022 \u4efb\u52d9\u6db5\u84cb\u56db\u7a2e\u7cfb\u7d71\u7a0b\u5f0f\u8a9e\u8a00\u53ca\u516b\u500b\u9818\u57df<br \/>\n\u2022 Qwen3.8-27B \u57fa\u6e96\u5206\u6578\u7531 34.87% \u63d0\u5347\u81f3 42.84%<br \/>\n\u2022 \u914d\u5408 Program-Agent \u5f8c\uff0c\u5206\u6578\u63d0\u5347\u81f3 55.71% \u6216 56.25%<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u6e2c\u8a66\u7d50\u679c\u986f\u793a\uff0cQwen3.8-27B \u914d\u5408 Program-Agent \u9054 55.71%\uff0cQwen3.8-27B-SFT \u5247\u9054 56.25%\uff0c\u8f03\u5404\u81ea\u57fa\u6e96\u660e\u986f\u63d0\u5347\uff0c\u4f46\u4ecd\u4f4e\u65bc Claude Opus 4.8 \u7684 70.9% \u548c GPT-5.6 Sol \u7684 69.9%\u3002GitHub \u5132\u5b58\u5eab\u4e3b\u8981\u662f\u8cc7\u8a0a\u4e2d\u5fc3\uff0c\u63d0\u4f9b\u8ad6\u6587\u3001\u9805\u76ee\u7db2\u7ad9\u53ca Hugging Face \u8cc7\u6e90\u9023\u7d50\uff1b\u539f\u59cb\u8cc7\u6599\u6c92\u6709\u4ea4\u4ee3\u5b8c\u6574\u5b89\u88dd\u6216\u672c\u5730\u57f7\u884c\u6b65\u9a5f\uff0c\u4f7f\u7528\u8005\u9700\u8981\u5148\u6309\u76f8\u95dc\u8cc7\u6e90\u4e86\u89e3\u53d6\u5f97\u65b9\u5f0f\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/programgym.github.io\/\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/Acm1011\/ProgramGym\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u7531\u53ef\u9a57\u8b49\u74b0\u5883\u5230\u6301\u7e8c\u958b\u767c\u6d41\u7a0b\uff0cProgramGym \u628a\u7a0b\u5f0f\u91cd\u88fd\u4ee3\u7406\u7684\u8a13\u7df4\u63a8\u5411\u9577\u6d41\u7a0b\u4efb\u52d9\u3002<\/p>\n","protected":false},"author":8,"featured_media":13265,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"wpai_generated_summary":"","wpai_meta_description":"","footnotes":""},"categories":[133,116,127,185,159,150,146,199],"tags":[],"class_list":["post-13266","post","type-post","status-publish","format-standard","hentry","category-133","category-agentic","category-127","category-qwen","category-vibe-coding","category-150","category-146","category-dataset-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/13266","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\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=13266"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/13266\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/13265"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=13266"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=13266"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=13266"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}