{"id":101106,"date":"2026-09-06T07:26:30","date_gmt":"2026-09-05T23:26:30","guid":{"rendered":"https:\/\/www.wsisp.com\/helps\/101106.html"},"modified":"2026-09-06T07:26:30","modified_gmt":"2026-09-05T23:26:30","slug":"terminal-universe-turning-agent-trajectories-into-scalable-terminal-environments-%e5%b0%86%e6%99%ba%e8%83%bd%e4%bd%93%e8%bd%a8%e8%bf%b9%e8%bd%ac%e5%8c%96%e4%b8%ba%e5%8f%af%e6%89%a9","status":"publish","type":"post","link":"https:\/\/www.wsisp.com\/helps\/101106.html","title":{"rendered":"Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments\u2014\u2014\u5c06\u667a\u80fd\u4f53\u8f68\u8ff9\u8f6c\u5316\u4e3a\u53ef\u6269\u5c55\u7684\u7ec8\u7aef\u73af\u5883"},"content":{"rendered":"<h3>\u4e00\u3001\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<\/h3>\n<h4>\u6838\u5fc3\u77db\u76fe<\/h4>\n<ul>\n<li>\n<p>\u8f68\u8ff9\u4e30\u5bcc&#xff0c;\u73af\u5883\u7a00\u7f3a&#xff1a;\u7ec8\u7aef\u4ee3\u7801\u667a\u80fd\u4f53\u4ea7\u751f\u5927\u91cf\u8f68\u8ff9\u6570\u636e&#xff0c;\u4f46\u771f\u5b9e\u3001\u53ef\u6267\u884c\u7684\u73af\u5883\u5374\u5f88\u5c11<\/p>\n<\/li>\n<li>\n<p>\u8f68\u8ff9 vs \u73af\u5883&#xff1a;\u8f68\u8ff9\u662f\u56fa\u5b9a\u7684\u5355\u6b21\u6f14\u793a&#xff0c;\u8d28\u91cf\u53d7\u9650\u4e8e\u751f\u6210\u6a21\u578b&#xff1b;\u73af\u5883\u53ef\u91cd\u590d\u67e5\u8be2\u3001\u9a8c\u8bc1\u3001\u751f\u6210\u66f4\u96be\u4efb\u52a1&#xff0c;\u662f\u540e\u8bad\u7ec3\u771f\u6b63\u9700\u8981\u7684\u8d44\u6e90<\/p>\n<\/li>\n<li>\n<p>\u4eba\u5de5\u6784\u5efa\u73af\u5883\u7684\u74f6\u9888&#xff1a;\u4e13\u5bb6\u6784\u5efa\u7684\u57fa\u51c6&#xff08;\u5982 Terminal-Bench&#xff09;\u8d28\u91cf\u9ad8\u4f46\u6570\u91cf\u6709\u9650&#xff0c;\u65e0\u6cd5\u6ee1\u8db3\u5927\u89c4\u6a21\u540e\u8bad\u7ec3\u9700\u6c42<\/p>\n<\/li>\n<\/ul>\n<h4>\u73b0\u6709\u65b9\u6cd5\u7684\u4e09\u6761\u8def\u5f84\u53ca\u5c40\u9650<\/h4>\n<table>\n<tr>\u8def\u5f84\u4ee3\u8868\u5de5\u4f5c\u5c40\u9650<\/tr>\n<tbody>\n<tr>\n<td>\u57fa\u4e8e\u4ee3\u7801\u4ed3\u5e93\u5386\u53f2<\/td>\n<td>SWE-Gym, R2E-Gym<\/td>\n<td>\u4f9d\u8d56\u73b0\u6709\u4ed3\u5e93&#xff0c;\u4efb\u52a1\u7c7b\u578b\u5355\u4e00&#xff08;\u4e3b\u8981\u662f\u4fee\u590d&#xff09;<\/td>\n<\/tr>\n<tr>\n<td>\u73af\u5883\u6270\u52a8<\/td>\n<td>SWE-smith, CLI-Gym<\/td>\n<td>\u4ec5\u9650\u4e8e\u4fee\u590d\u7c7b\u4efb\u52a1&#xff0c;\u6ce8\u5165\u9519\u8bef\u7c7b\u578b\u6709\u9650<\/td>\n<\/tr>\n<tr>\n<td>\u4efb\u52a1\u6761\u4ef6\u5408\u6210<\/td>\n<td>Endless Terminal, TMax<\/td>\n<td>\u751f\u6210\u7684\u73af\u5883\u4e0d\u591f\u771f\u5b9e&#xff0c;\u901a\u5e38\u662f&#034;\u6574\u6d01\u7684\u5c0f\u578b\u9879\u76ee&#034;<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u5173\u952e\u6d1e\u5bdf&#xff1a;\u8f68\u8ff9\u672c\u8eab\u8bb0\u5f55\u4e86\u73af\u5883\u7684\u6587\u4ef6\u7ed3\u6784\u3001\u5185\u5bb9\u53d8\u5316&#xff0c;\u8db3\u4ee5\u53cd\u5411\u91cd\u5efa\u53ef\u6267\u884c\u73af\u5883\u2014\u2014\u8fd9\u6b63\u662f Terminal-Universe \u7684\u51fa\u53d1\u70b9\u3002<\/p>\n<h3>\u4e8c\u3001\u6838\u5fc3\u65b9\u6cd5<\/h3>\n<h4>\u6574\u4f53\u601d\u8def<\/h4>\n<p>\u5c06&#034;\u4ece\u73af\u5883\u751f\u6210\u8f68\u8ff9&#034;\u7684\u4f20\u7edf\u6d41\u7a0b\u53cd\u5411&#xff1a;\u4ece\u8f68\u8ff9\u6062\u590d\u73af\u5883 \u2192 \u5728\u73af\u5883\u4e2d\u5408\u6210\u65b0\u4efb\u52a1 \u2192 \u7528\u6559\u5e08\u6a21\u578b\u751f\u6210\u89e3\u51b3\u65b9\u6848\u8f68\u8ff9 \u2192 SFT \u8bad\u7ec3<\/p>\n<h4>2.1 \u73af\u5883\u91cd\u5efa&#xff08;\u4e09\u9636\u6bb5&#xff09;<\/h4>\n<p>\u9636\u6bb5\u4e00&#xff1a;\u786e\u5b9a\u6027\u56de\u653e<\/p>\n<ul>\n<li>\n<p>\u6309\u65f6\u95f4\u987a\u5e8f\u56de\u653e\u6587\u4ef6\u8bfb\u5199\u64cd\u4f5c<\/p>\n<\/li>\n<li>\n<p>\u6062\u590d\u6bcf\u4e2a\u6587\u4ef6\u5728\u667a\u80fd\u4f53\u4fee\u6539\u4e4b\u524d\u7684\u72b6\u6001<\/p>\n<\/li>\n<li>\n<p>\u6392\u9664\u667a\u80fd\u4f53\u521b\u5efa\u7684\u6587\u4ef6\u548c\u6240\u6709\u540e\u671f\u66f4\u6539<\/p>\n<\/li>\n<li>\n<p>\u7ed3\u679c&#xff1a;\u90e8\u5206\u5de5\u4f5c\u7a7a\u95f4&#xff08;\u53ea\u5305\u542b\u8f68\u8ff9\u4e2d\u66b4\u9732\u7684\u6587\u4ef6&#xff09;<\/p>\n<\/li>\n<\/ul>\n<p>\u9636\u6bb5\u4e8c&#xff1a;\u667a\u80fd\u4f53\u8865\u5168<\/p>\n<ul>\n<li>\n<p>\u8865\u5168\u667a\u80fd\u4f53\u6839\u636e\u4efb\u52a1\u548c\u73b0\u6709\u6587\u4ef6&#xff0c;\u586b\u5145\u7f3a\u5931\u5185\u5bb9&#xff1a;<\/p>\n<ul>\n<li>\n<p>\u521b\u5efa\u7f3a\u5931\u7684\u914d\u7f6e\u6587\u4ef6\u3001\u811a\u672c\u3001\u6570\u636e\u3001\u6a21\u5f0f\u6587\u4ef6<\/p>\n<\/li>\n<li>\n<p>\u8865\u5168\u88ab\u622a\u65ad\u7684\u6587\u4ef6<\/p>\n<\/li>\n<li>\n<p>\u5b89\u88c5\/\u914d\u7f6e\u4f9d\u8d56\u9879<\/p>\n<\/li>\n<\/ul>\n<\/li>\n<li>\n<p>\u5173\u952e\u7ea6\u675f&#xff1a;\u4e0d\u5b9e\u73b0\u4efb\u52a1\u672c\u8eab&#xff0c;\u53ea\u63d0\u4f9b\u53ef\u5de5\u4f5c\u7684\u4e0a\u4e0b\u6587<\/p>\n<\/li>\n<\/ul>\n<p>\u9636\u6bb5\u4e09&#xff1a;\u73af\u5883\u8fc7\u6ee4<\/p>\n<ul>\n<li>\n<p>\u667a\u80fd\u4f53\u8bc4\u5224\u5458\u68c0\u67e5\u5de5\u4f5c\u7a7a\u95f4\u662f\u5426&#034;\u4efb\u52a1\u5145\u5206&#034;<\/p>\n<\/li>\n<li>\n<p>\u5145\u5206\u6807\u51c6&#xff1a;\u6e90\u6587\u4ef6\u3001\u914d\u7f6e\u3001\u6570\u636e\u7ed3\u6784\u8db3\u4ee5\u8ba9\u6709\u80fd\u529b\u667a\u80fd\u4f53\u5f00\u59cb\u5de5\u4f5c<\/p>\n<\/li>\n<li>\n<p>\u7ed3\u679c&#xff1a;37,273 \u4e2a\u5145\u5206\u73af\u5883&#xff08;\u7ec8\u7aef\u6c60 93.5%&#xff0c;SWE \u6c60 77.1%&#xff09;<\/p>\n<\/li>\n<\/ul>\n<h4>2.2 \u56db\u79cd\u91cd\u65b0\u67e5\u8be2\u673a\u5236<\/h4>\n<p>\u2460 \u610f\u56fe\u6062\u590d<\/p>\n<ul>\n<li>\n<p>\u4ece\u6e90\u8f68\u8ff9\u53cd\u5411\u63d0\u53d6\u7528\u6237\u7684\u539f\u59cb\u610f\u56fe<\/p>\n<\/li>\n<li>\n<p>\u5355\u8f6e \u2192 \u76f4\u63a5\u4f7f\u7528&#xff1b;\u591a\u8f6e \u2192 \u5408\u5e76\u6f84\u6e05\u548c\u7ea6\u675f<\/p>\n<\/li>\n<li>\n<p>\u6062\u590d\u51fa&#xff1a;\u6838\u5fc3\u76ee\u6807\u3001\u6210\u529f\u6807\u51c6\u3001\u7ea6\u675f\u6761\u4ef6<\/p>\n<\/li>\n<\/ul>\n<p>\u2461 \u5355\u5de5\u4f5c\u7a7a\u95f4\u5408\u6210<\/p>\n<ul>\n<li>\n<p>\u5728\u5355\u4e2a\u91cd\u5efa\u7684\u5de5\u4f5c\u7a7a\u95f4\u5185\u63a2\u7d22<\/p>\n<\/li>\n<li>\n<p>\u6bcf\u4e2a\u73af\u5883\u751f\u6210 5 \u4e2a\u5019\u9009\u4efb\u52a1&#xff08;\u57fa\u4e8e\u771f\u5b9e\u573a\u666f\u3001\u53ef\u9a8c\u8bc1\u3001\u591a\u6837\u5316&#xff09;<\/p>\n<\/li>\n<li>\n<p>\u968f\u673a\u9009 1 \u4e2a\u8fdb\u884c\u5c55\u5f00\u548c\u9a8c\u8bc1<\/p>\n<\/li>\n<li>\n<p>\u4ea7\u51fa&#xff1a;25,386 \u6761\u8f68\u8ff9<\/p>\n<\/li>\n<\/ul>\n<p>\u2462 \u5e7f\u5ea6\u6269\u5c55&#xff1a;\u8de8\u5de5\u4f5c\u7a7a\u95f4\u5408\u6210<\/p>\n<ul>\n<li>\n<p>\u6316\u6398\u4e0d\u540c\u5de5\u4f5c\u7a7a\u95f4\u95f4\u7684\u65b9\u5411\u6027\u4f9d\u8d56\u5173\u7cfb<\/p>\n<\/li>\n<li>\n<p>\u6d41\u7a0b&#xff1a;<\/p>\n<ul>\n<li>\n<p>\u5206\u6790\u6bcf\u4e2a\u5de5\u4f5c\u7a7a\u95f4\u7684\u9886\u57df\u548c\u80fd\u529b<\/p>\n<\/li>\n<li>\n<p>TF-IDF \u68c0\u7d22\u76f8\u4f3c\u5019\u9009\u5bf9<\/p>\n<\/li>\n<li>\n<p>LLM \u5224\u65ad\u4f9d\u8d56\u65b9\u5411&#xff08;\u53c2\u8003\u6709&#xff0c;\u76ee\u6807\u7f3a&#xff09;<\/p>\n<\/li>\n<li>\n<p>\u751f\u6210\u8de8\u4ee3\u7801\u5e93\u4efb\u52a1&#xff08;\u76ee\u6807\u53ef\u5199 &#043; \u53c2\u8003\u53ea\u8bfb&#xff09;<\/p>\n<\/li>\n<\/ul>\n<\/li>\n<li>\n<p>\u4efb\u52a1\u7279\u70b9&#xff1a;\u6c42\u89e3\u8005\u5fc5\u987b\u81ea\u5df1\u9605\u8bfb\u53c2\u8003\u4ed3\u5e93\u3001\u5185\u5316\u5b9e\u73b0\u3001\u9002\u914d\u5230\u76ee\u6807<\/p>\n<\/li>\n<li>\n<p>\u6559\u5e08 pass&#064;1 \u4ece 72.3%&#xff08;\u5355\u5de5\u4f5c\u7a7a\u95f4&#xff09;\u964d\u81f3 49.2%&#xff08;\u8de8\u5de5\u4f5c\u7a7a\u95f4&#xff09;&#xff0c;\u96be\u5ea6\u663e\u8457\u66f4\u9ad8<\/p>\n<\/li>\n<li>\n<p>\u4ea7\u51fa&#xff1a;3,512 \u6761\u8f68\u8ff9<\/p>\n<\/li>\n<\/ul>\n<p>\u2463 \u6df1\u5ea6\u6269\u5c55&#xff1a;\u591a\u8f6e\u7528\u6237\u67e5\u8be2<\/p>\n<ul>\n<li>\n<p>\u5df2\u89e3\u51b3\u4efb\u52a1\u57fa\u7840\u4e0a\u5f15\u5165\u7528\u6237\u667a\u80fd\u4f53<\/p>\n<\/li>\n<li>\n<p>\u4e09\u79cd\u4ea4\u4e92\u98ce\u683c&#xff1a;<\/p>\n<ul>\n<li>\n<p>\u7279\u6027\u6269\u5c55&#xff08;62.7%&#xff09;&#xff1a;\u901a\u8fc7\u540e\u52a0\u65b0\u9700\u6c42<\/p>\n<\/li>\n<li>\n<p>\u7279\u6027\u4fee\u8ba2&#xff08;29.3%&#xff09;&#xff1a;\u5931\u8d25\u540e\u8981\u6c42\u4fee\u590d<\/p>\n<\/li>\n<li>\n<p>\u7279\u6027\u51b2\u7a81&#xff08;8.0%&#xff09;&#xff1a;\u6709\u610f\u8986\u76d6\u4e4b\u524d\u7684\u9700\u6c42<\/p>\n<\/li>\n<\/ul>\n<\/li>\n<li>\n<p>\u6838\u5fc3\u673a\u5236&#xff1a;<\/p>\n<ul>\n<li>\n<p>\u9700\u6c42\u8ddf\u8e2a\u5668\u8bb0\u5f55\u6d3b\u8dc3\/\u5df2\u6ee1\u8db3\/\u5df2\u66f4\u65b0\u9700\u6c42<\/p>\n<\/li>\n<li>\n<p>\u8f6e\u6b21\u7ea7\u9a8c\u8bc1&#xff08;\u79c1\u6709\u9a8c\u8bc1\u5668&#xff0c;\u6c42\u89e3\u8005\u770b\u4e0d\u5230\u6d4b\u8bd5&#xff09;<\/p>\n<\/li>\n<li>\n<p>\u5931\u8d25\u2192\u81ea\u7136\u8bed\u8a00\u62b1\u6028&#xff0c;\u4fdd\u7559\u9519\u8bef\u8bca\u65ad\u548c\u6062\u590d\u76d1\u7763<\/p>\n<\/li>\n<\/ul>\n<\/li>\n<li>\n<p>\u5e73\u5747 4.51 \u8f6e&#xff0c;69.6% \u5305\u542b\u53ef\u4fee\u590d\u7684\u5931\u8d25<\/p>\n<\/li>\n<li>\n<p>\u4ea7\u51fa&#xff1a;3,079 \u6761\u8f68\u8ff9<\/p>\n<\/li>\n<\/ul>\n<h4>2.3 \u9a8c\u8bc1\u4e0e\u8fc7\u6ee4<\/h4>\n<ul>\n<li>\n<p>\u6bcf\u4e2a\u4efb\u52a1\u914d\u5907\u667a\u80fd\u4f53\u7f16\u5199\u7684\u00a0pytest \u9a8c\u8bc1\u5668<\/p>\n<\/li>\n<li>\n<p>\u8981\u6c42&#xff1a;\u65b0\u529f\u80fd\u6d4b\u8bd5\u5728\u521d\u59cb\u72b6\u6001\u5fc5\u987b\u5931\u8d25&#xff08;&#034;\u7ea2\u68c0&#034;&#xff09;&#xff0c;\u4fdd\u7559\u6d4b\u8bd5\u5fc5\u987b\u901a\u8fc7<\/p>\n<\/li>\n<li>\n<p>\u6559\u5e08\u6a21\u578b\u5c55\u5f00\u89e3\u51b3\u65b9\u6848&#xff08;Qwen3.7-Max&#xff0c;Claude Code \u6846\u67b6&#xff09;<\/p>\n<\/li>\n<li>\n<p>\u4ec5\u4fdd\u7559\u6240\u6709\u6d4b\u8bd5\u901a\u8fc7\u7684\u8f68\u8ff9<\/p>\n<\/li>\n<li>\n<p>\u603b SFT \u6570\u636e&#xff1a;31,977 \u6761\u8bb0\u5f55&#xff0c;\u7ea6 14.2 \u4ebf token<\/p>\n<\/li>\n<\/ul>\n<h3>\u4e09\u3001\u5b9e\u9a8c\u4e0e\u7ed3\u679c<\/h3>\n<h4>3.1 \u5b9e\u9a8c\u8bbe\u7f6e<\/h4>\n<ul>\n<li>\n<p>\u57fa\u7840\u6a21\u578b&#xff1a;Qwen3.5-27B<\/p>\n<\/li>\n<li>\n<p>\u8bad\u7ec3&#xff1a;SFT \u4e24\u8f6e&#xff0c;\u5b66\u4e60\u7387 7\u00d710\u207b\u2076&#xff0c;\u6279\u6b21 256&#xff0c;\u5e8f\u5217\u957f\u5ea6 256k<\/p>\n<\/li>\n<li>\n<p>\u5355\u8f6e\u8bc4\u4f30&#xff1a;Terminal-Bench 2.0\/2.1&#xff08;Terminus2-XML \u548c Claude Code&#xff09;<\/p>\n<\/li>\n<li>\n<p>\u591a\u8f6e\u8bc4\u4f30&#xff1a;EvoCode-Bench v2&#xff08;MT&#064;4 \u548c Case Score&#xff09;<\/p>\n<\/li>\n<\/ul>\n<h4>3.2 \u4e3b\u8981\u7ed3\u679c<\/h4>\n<p>\u5355\u8f6e\u6027\u80fd&#xff08;Terminal-Bench 2.1&#xff09;<\/p>\n<table>\n<tr>\u6a21\u578b\u6570\u636e\u91cf\u5206\u6570<\/tr>\n<tbody>\n<tr>\n<td>Qwen3.5-27B&#xff08;\u57fa\u7840&#xff09;<\/td>\n<td>&#8211;<\/td>\n<td>46.2%<\/td>\n<\/tr>\n<tr>\n<td>Terminal-Universe-27B<\/td>\n<td>32k<\/td>\n<td>58.1%\u00a0(&#043;11.9)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u591a\u8f6e\u6027\u80fd&#xff08;EvoCode-Bench v2&#xff09;<\/p>\n<table>\n<tr>\u6a21\u578bMT&#064;4Case Score<\/tr>\n<tbody>\n<tr>\n<td>Qwen3.5-27B&#xff08;\u57fa\u7840&#xff09;<\/td>\n<td>6.3%<\/td>\n<td>67.8%<\/td>\n<\/tr>\n<tr>\n<td>Terminal-Universe-27B<\/td>\n<td>20.1%\u00a0(&#043;13.8)<\/td>\n<td>76.1%\u00a0(&#043;8.3)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u5bf9\u6bd4\u5176\u4ed6\u4efb\u52a1\u5408\u6210\u65b9\u6cd5&#xff1a;Terminal-Universe \u5728\u6240\u6709\u5bf9\u6bd4\u65b9\u6cd5\u4e2d\u8868\u73b0\u6700\u4f73&#xff08;\u8868 3&#xff09;\u3002<\/p>\n<h4>3.3 \u5173\u952e\u6d88\u878d\u7ed3\u8bba<\/h4>\n<table>\n<tr>\u6d88\u878d\u95ee\u9898\u6838\u5fc3\u53d1\u73b0<\/tr>\n<tbody>\n<tr>\n<td>\u91cd\u5efa\u662f\u5426\u5fc5\u8981&#xff1f;<\/td>\n<td>\u610f\u56fe\u6062\u590d&#xff08;52.1%&#xff09;\u663e\u8457\u4f18\u4e8e\u76f4\u63a5\u6a21\u4eff\u6e90\u8f68\u8ff9 SFT&#xff08;36.7%&#xff09;&#xff0c;\u5dee\u8ddd 15.4 \u4e2a\u767e\u5206\u70b9<\/td>\n<\/tr>\n<tr>\n<td>\u667a\u80fd\u4f53\u8865\u5168\u662f\u5426\u91cd\u8981&#xff1f;<\/td>\n<td>\u8865\u5168\u540e 52.9% vs \u4ec5\u56de\u653e 48.7%&#xff0c;\u5dee\u8ddd 4.2 \u4e2a\u767e\u5206\u70b9&#xff1b;\u8865\u5168\u628a\u5145\u5206\u7387\u4ece 40.2% \u63d0\u5347\u5230 93.5%<\/td>\n<\/tr>\n<tr>\n<td>\u9a8c\u8bc1\u5668\u8fc7\u6ee4\u662f\u5426\u6709\u7528&#xff1f;<\/td>\n<td>\u5bf9\u8de8\u5de5\u4f5c\u7a7a\u95f4\u5c24\u5176\u5173\u952e&#xff1a;\u8fc7\u6ee4\u540e 55.4% vs \u4e0d\u8fc7\u6ee4 53.2%&#xff0c;\u4e14\u6570\u636e\u91cf\u51cf\u534a<\/td>\n<\/tr>\n<tr>\n<td>\u5e7f\u5ea6&#xff08;\u8de8\u5de5\u4f5c\u7a7a\u95f4&#xff09;\u6709\u5e2e\u52a9\u5417&#xff1f;<\/td>\n<td>\u5355\u5de5\u4f5c\u7a7a\u95f4 56.4% \u2192 \u52a0\u8de8\u5de5\u4f5c\u7a7a\u95f4 58.4%&#xff0c;\u63d0\u5347 2 \u4e2a\u767e\u5206\u70b9<\/td>\n<\/tr>\n<tr>\n<td>\u6df1\u5ea6&#xff08;\u591a\u8f6e&#xff09;\u6709\u5e2e\u52a9\u5417&#xff1f;<\/td>\n<td>\u5355\u5de5\u4f5c\u7a7a\u95f4 MT&#064;4 18.4 \u2192 \u52a0\u591a\u8f6e 21.0&#xff0c;\u6848\u4f8b\u5206\u6570 71.9 \u2192 76.9<\/td>\n<\/tr>\n<tr>\n<td>\u66f4\u591a\u73af\u5883 vs \u6bcf\u73af\u5883\u66f4\u591a\u67e5\u8be2&#xff1f;<\/td>\n<td>\u56fa\u5b9a\u8bb0\u5f55\u6570\u4e0b&#xff1a;25k \u4e2a\u73af\u5883\u5404 1 \u67e5\u8be2&#xff08;56.4%&#xff09;&gt; 5k \u4e2a\u73af\u5883\u5404 5 \u67e5\u8be2&#xff08;55.0%&#xff09;<\/td>\n<\/tr>\n<tr>\n<td>\u80fd\u5426\u6cdb\u5316\u5230 SWE&#xff1f;<\/td>\n<td>SWE \u6c60\u5355\u72ec\u8bad\u7ec3\u8fbe 48.6%&#xff08;\u57fa\u7840 46.2%&#xff09;&#xff0c;\u6709\u6548\u4f46\u5f31\u4e8e\u7ec8\u7aef\u6c60&#xff08;52.9%&#xff09;<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h3>\u56db\u3001\u6570\u636e\u7edf\u8ba1<\/h3>\n<h4>\u6e90\u6570\u636e\u4e0e\u91cd\u5efa\u89c4\u6a21<\/h4>\n<table>\n<tr>\u6e90\u8bed\u6599\u5e93\u7c7b\u578b\u539f\u59cb\u8f68\u8ff9\u91cd\u5efa\u73af\u5883<\/tr>\n<tbody>\n<tr>\n<td>LFM2-Terminal<\/td>\n<td>\u7ec8\u7aef<\/td>\n<td>139,841<\/td>\n<td>46,037<\/td>\n<\/tr>\n<tr>\n<td>SWE-rebench<\/td>\n<td>SWE<\/td>\n<td>67,074<\/td>\n<td>6,118<\/td>\n<\/tr>\n<tr>\n<td>SWE-smith<\/td>\n<td>SWE<\/td>\n<td>95,851<\/td>\n<td>8,476<\/td>\n<\/tr>\n<tr>\n<td>CoderForge<\/td>\n<td>SWE<\/td>\n<td>32,964<\/td>\n<td>3,978<\/td>\n<\/tr>\n<tr>\n<td>\u5176\u4ed6<\/td>\n<td>&#8211;<\/td>\n<td>~24k<\/td>\n<td>~3.6k<\/td>\n<\/tr>\n<tr>\n<td>\u603b\u8ba1<\/td>\n<td><\/td>\n<td>359,593<\/td>\n<td>68,263<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h4>\u6700\u7ec8 SFT \u8bed\u6599<\/h4>\n<table>\n<tr>\u7c7b\u578b\u8bb0\u5f55\u6570\u4e2d\u4f4d\u6570\u8f6e\u6b21\u4e2d\u4f4d\u6570\u5de5\u5177\u8c03\u7528\u4e2d\u4f4d\u6570 Token<\/tr>\n<tbody>\n<tr>\n<td>\u610f\u56fe\u6062\u590d<\/td>\n<td>35,809<\/td>\n<td>12<\/td>\n<td>17<\/td>\n<td>36.1k<\/td>\n<\/tr>\n<tr>\n<td>\u5355\u5de5\u4f5c\u7a7a\u95f4<\/td>\n<td>25,386<\/td>\n<td>14<\/td>\n<td>20<\/td>\n<td>30.4k<\/td>\n<\/tr>\n<tr>\n<td>\u8de8\u5de5\u4f5c\u7a7a\u95f4<\/td>\n<td>3,512<\/td>\n<td>23<\/td>\n<td>38<\/td>\n<td>46.5k<\/td>\n<\/tr>\n<tr>\n<td>\u591a\u8f6e<\/td>\n<td>3,079<\/td>\n<td>92<\/td>\n<td>102<\/td>\n<td>126.1k<\/td>\n<\/tr>\n<tr>\n<td>\u5408\u8ba1<\/td>\n<td>31,977&#xff08;\u9a8c\u8bc1\u540e&#xff09;<\/td>\n<td>&#8211;<\/td>\n<td>&#8211;<\/td>\n<td>~1.42B<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h3>\u4e94\u3001\u4e3b\u8981\u8d21\u732e<\/h3>\n<li>\n<p>\u65b0\u89c6\u89d2&#xff1a;\u9996\u6b21\u7cfb\u7edf\u6027\u5730\u5c06\u667a\u80fd\u4f53\u8f68\u8ff9\u53cd\u5411\u91cd\u5efa\u4e3a\u53ef\u590d\u7528\u73af\u5883&#xff0c;\u800c\u975e\u4ece\u5934\u751f\u6210<\/p>\n<\/li>\n<li>\n<p>\u5b8c\u6574\u6846\u67b6&#xff1a;<\/p>\n<ul>\n<li>\n<p>\u786e\u5b9a\u6027\u56de\u653e &#043; \u667a\u80fd\u4f53\u8865\u5168\u7684\u73af\u5883\u91cd\u5efa<\/p>\n<\/li>\n<li>\n<p>\u56db\u79cd\u91cd\u65b0\u67e5\u8be2\u673a\u5236&#xff08;\u610f\u56fe\u6062\u590d\u3001\u5355\u5de5\u4f5c\u7a7a\u95f4\u3001\u8de8\u5de5\u4f5c\u7a7a\u95f4\u3001\u591a\u8f6e&#xff09;<\/p>\n<\/li>\n<li>\n<p>\u81ea\u52a8\u5316\u9a8c\u8bc1\u548c\u8fc7\u6ee4<\/p>\n<\/li>\n<\/ul>\n<\/li>\n<li>\n<p>\u89c4\u6a21\u4e0e\u5b9e\u8bc1&#xff1a;<\/p>\n<ul>\n<li>\n<p>37.3k \u4efb\u52a1\u5145\u5206\u73af\u5883<\/p>\n<\/li>\n<li>\n<p>32k \u9a8c\u8bc1\u901a\u8fc7\u7684\u9ad8\u8d28\u91cf SFT \u8f68\u8ff9<\/p>\n<\/li>\n<li>\n<p>\u5355\u8f6e &#043;11.9 \u70b9&#xff0c;\u591a\u8f6e &#043;13.8 \u70b9\u7684\u663e\u8457\u63d0\u5347<\/p>\n<\/li>\n<\/ul>\n<\/li>\n<li>\n<p>\u7cfb\u7edf\u6d88\u878d&#xff1a;\u5168\u9762\u9a8c\u8bc1\u6bcf\u4e2a\u8bbe\u8ba1\u51b3\u7b56\u7684\u5fc5\u8981\u6027\u548c\u8d21\u732e<\/p>\n<\/li>\n<h3>\u516d\u3001\u5c40\u9650\u6027\u4e0e\u672a\u6765\u65b9\u5411<\/h3>\n<table>\n<tr>\u5c40\u9650\u6027\u672a\u6765\u6539\u8fdb\u65b9\u5411<\/tr>\n<tbody>\n<tr>\n<td>\u7edf\u4e00\u4f7f\u7528 Ubuntu 24.04 \u5bb9\u5668&#xff0c;\u672a\u9488\u5bf9\u4ed3\u5e93\u5b9a\u5236<\/td>\n<td>\u96c6\u6210 SWE-Factory \u7b49\u81ea\u52a8\u5316\u73af\u5883\u6784\u5efa<\/td>\n<\/tr>\n<tr>\n<td>\u9886\u57df\/\u8bed\u8a00\u5206\u5e03\u53d7\u9650\u4e8e\u6e90\u8f68\u8ff9\u8986\u76d6<\/td>\n<td>\u6269\u5c55\u66f4\u591a\u6e90\u8f68\u8ff9\u7c7b\u578b\u548c\u9886\u57df<\/td>\n<\/tr>\n<tr>\n<td>\u5355\u6559\u5e08\u751f\u6210\u4efb\u52a1\u3001\u89e3\u6cd5\u548c\u9a8c\u8bc1\u5668<\/td>\n<td>\u591a\u6559\u5e08 &#043; \u72ec\u7acb\u9a8c\u8bc1\u5668\u6a21\u578b<\/td>\n<\/tr>\n<tr>\n<td>\u672a\u6df7\u5408\u7ec8\u7aef\u6c60\u548c SWE \u6c60<\/td>\n<td>\u6df7\u5408\u8bad\u7ec3\u53ef\u80fd\u8fdb\u4e00\u6b65\u63d0\u5347 SWE \u6027\u80fd<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>Terminal-Universe \u901a\u8fc7\u4ece\u5df2\u6709\u667a\u80fd\u4f53\u8f68\u8ff9\u4e2d\u53cd\u5411\u6062\u590d\u53ef\u6267\u884c\u73af\u5883&#xff0c;\u518d\u5728\u8fd9\u4e9b\u73af\u5883\u4e2d\u5408\u6210\u65b0\u4efb\u52a1\u5e76\u7528\u5f3a\u6559\u5e08\u751f\u6210\u9ad8\u8d28\u91cf\u6f14\u793a&#xff0c;\u5b9e\u73b0\u4e86\u7ec8\u7aef\u667a\u80fd\u4f53\u8bad\u7ec3\u6570\u636e\u7684\u89c4\u6a21\u5316\u3001\u9ad8\u8d28\u91cf\u3001\u53ef\u9a8c\u8bc1\u6784\u5efa&#xff0c;\u5728\u5355\u8f6e\u548c\u591a\u8f6e\u57fa\u51c6\u4e0a\u5747\u53d6\u5f97\u663e\u8457\u63d0\u5347\u3002<\/p>\n<p>\u8fd9\u91cc\u662f\u81ea\u5df1\u7684\u8bba\u6587\u9605\u8bfb\u8bb0\u5f55&#xff0c;\u611f\u5174\u8da3\u7684\u8bdd\u53ef\u4ee5\u53c2\u8003\u4e00\u4e0b&#xff0c;\u5982\u679c\u9700\u8981\u9605\u8bfb\u539f\u6587\u7684\u8bdd\u53ef\u4ee5\u770b\u8fd9\u91cc&#xff0c;\u5982\u4e0b\u6240\u793a&#xff1a;<\/p>\n<p class=\"img-center\"><img loading=\"lazy\" decoding=\"async\" alt=\"\" height=\"1614\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260905232623-6a9ca51f9477e.png\" width=\"2512\" \/><\/p>\n<h3>\u6458\u8981<\/h3>\n<p>\u968f\u7740\u57fa\u4e8e\u7ec8\u7aef\u7684\u4ee3\u7801\u667a\u80fd\u4f53\u65e5\u76ca\u666e\u53ca&#xff0c;\u76f8\u5e94\u7684\u667a\u80fd\u4f53\u8f68\u8ff9\u5df2\u5927\u89c4\u6a21\u79ef\u7d2f&#xff0c;\u800c\u771f\u5b9e\u7684\u3001\u53ef\u6267\u884c\u7684\u73af\u5883\u4ecd\u7136\u7a00\u7f3a\u3002\u7136\u800c&#xff0c;\u73af\u5883\u624d\u662f\u667a\u80fd\u4f53\u540e\u8bad\u7ec3\u771f\u6b63\u9700\u8981\u7684\u8d44\u6e90&#xff1a;\u6bcf\u4e2a\u73af\u5883\u90fd\u53ef\u4ee5\u88ab\u91cd\u65b0\u67e5\u8be2\u4ee5\u751f\u6210\u8bb8\u591a\u53ef\u9a8c\u8bc1\u7684\u4efb\u52a1&#xff0c;\u5e76\u63d0\u4f9b\u6267\u884c\u53cd\u9988&#xff0c;\u800c\u8f68\u8ff9\u53ea\u662f\u4e00\u4e2a\u56fa\u5b9a\u7684\u6f14\u793a\u3002\u6211\u4eec\u89c2\u5bdf\u5230&#xff0c;\u4e0e\u5176\u4ece\u5934\u751f\u6210\u73af\u5883&#xff0c;\u4e0d\u5982\u5229\u7528\u73b0\u6709\u667a\u80fd\u4f53\u8f68\u8ff9\u4e2d\u7684\u5de5\u5177\u6267\u884c\u5386\u53f2&#xff0c;\u8fd9\u4e9b\u5386\u53f2\u63ed\u793a\u4e86\u5b83\u4eec\u8fd0\u884c\u6240\u5728\u73af\u5883\u7684\u7ed3\u6784\u548c\u5185\u5bb9&#xff0c;\u8fd9\u4f7f\u5f97\u4ece\u8f68\u8ff9\u672c\u8eab\u91cd\u5efa\u8fd9\u4e9b\u73af\u5883\u6210\u4e3a\u53ef\u80fd\u3002\u56e0\u6b64&#xff0c;\u6211\u4eec\u5f15\u5165\u4e86 Terminal-Universe&#xff0c;\u4e00\u4e2a\u6846\u67b6&#xff0c;\u5b83\u5c06\u6bcf\u6761\u8f68\u8ff9\u8f6c\u5316\u4e3a\u4e00\u4e2a\u53ef\u91cd\u7528\u7684\u73af\u5883&#xff0c;\u5e76\u5bf9\u5176\u8fdb\u884c\u63a2\u7d22\u4ee5\u5408\u6210\u65b0\u4efb\u52a1\u548c\u8fdb\u884c\u6301\u7eed\u4ea4\u4e92\u3002\u5177\u4f53\u6765\u8bf4&#xff0c;Terminal-Universe \u56de\u653e\u8f68\u8ff9\u4e2d\u8bb0\u5f55\u7684\u6587\u4ef6\u64cd\u4f5c&#xff0c;\u5c06\u6bcf\u4e2a\u6587\u4ef6\u6062\u590d\u5230\u667a\u80fd\u4f53\u4fee\u6539\u4e4b\u524d\u7684\u72b6\u6001&#xff0c;\u4ece\u800c\u4ea7\u751f\u4e00\u4e2a\u90e8\u5206\u5de5\u4f5c\u7a7a\u95f4&#xff1b;\u7136\u540e&#xff0c;\u4e00\u4e2a\u8865\u5168\u667a\u80fd\u4f53\u63d0\u4f9b\u7f3a\u5931\u7684\u6587\u4ef6\u548c\u4f9d\u8d56\u9879\u3002\u5728\u8fd9\u4e2a\u6062\u590d\u7684\u5de5\u4f5c\u7a7a\u95f4\u4e0a&#xff0c;\u6211\u4eec\u65e2\u91cd\u5efa\u4e86\u539f\u59cb\u610f\u56fe\u4efb\u52a1&#xff0c;\u4e5f\u5408\u6210\u4e86\u5168\u65b0\u7684\u4efb\u52a1\u3002\u6b64\u5916&#xff0c;\u6211\u4eec\u8fd8\u6cbf\u7740\u4e24\u4e2a\u4e92\u8865\u7684\u7ef4\u5ea6\u6269\u5c55\u73af\u5883\u5bf9\u5e94\u7684\u4efb\u52a1&#xff1a;\u5e7f\u5ea6&#xff08;breadth&#xff09;\u548c\u6df1\u5ea6&#xff08;depth&#xff09;&#xff0c;\u4ee5\u590d\u73b0\u771f\u5b9e\u5de5\u7a0b\u5b9e\u8df5\u4e2d\u7684\u5e38\u89c4\u6a21\u5f0f\u3002\u5bf9\u4e8e\u5e7f\u5ea6&#xff0c;\u6211\u4eec\u6316\u6398\u76f8\u5173\u73af\u5883\u4e4b\u95f4\u7684\u65b9\u5411\u6027\u4f9d\u8d56\u5173\u7cfb&#xff0c;\u5e76\u5408\u6210\u8de8\u8d8a\u591a\u4e2a\u4ee3\u7801\u5e93\u7684\u8de8\u5de5\u4f5c\u7a7a\u95f4\u67e5\u8be2&#xff0c;\u6b63\u5982\u5f00\u53d1\u8005\u5728\u73b0\u5b9e\u4e16\u754c\u5f00\u53d1\u4e2d\u7ecf\u5e38\u505a\u7684\u90a3\u6837\u3002\u5bf9\u4e8e\u6df1\u5ea6&#xff0c;\u6211\u4eec\u5c06\u521d\u59cb\u7684\u5355\u8f6e\u67e5\u8be2\u6269\u5c55\u4e3a\u591a\u8f6e\u4f1a\u8bdd&#xff0c;\u901a\u8fc7\u4e00\u4e2a\u7528\u6237\u667a\u80fd\u4f53\u6765\u6355\u6349\u8fed\u4ee3\u7684\u7528\u6237\u53cd\u9988\u548c\u9700\u6c42\u7ec6\u5316\u3002\u5c06 Terminal-Universe \u5e94\u7528\u4e8e\u516c\u5f00\u53ef\u7528\u7684\u7ec8\u7aef\u667a\u80fd\u4f53\u8f68\u8ff9&#xff0c;\u4ea7\u751f\u4e86 37.3k \u4e2a\u4efb\u52a1\u5145\u5206\u7684\u73af\u5883\u3002\u5728\u8be5\u8bed\u6599\u5e93\u4e0a\u5bf9 Qwen3.5-27B \u8fdb\u884c\u76d1\u7763\u5fae\u8c03&#xff0c;\u5728 Terminal-Bench 2.1 \u4e0a\u5c06\u5355\u8f6e\u6027\u80fd\u63d0\u5347\u4e86 11.9 \u4e2a\u70b9&#xff0c;\u5728 EvoCode-Bench v2 MT&#064;4 \u4e0a\u5c06\u591a\u8f6e\u6027\u80fd\u63d0\u5347\u4e86 13.8 \u4e2a\u70b9\u3002<\/p>\n<h3>1 \u5f15\u8a00<\/h3>\n<p>\u968f\u7740\u57fa\u4e8e\u7ec8\u7aef\u7684\u4ee3\u7801\u667a\u80fd\u4f53\u65e5\u76ca\u666e\u53ca&#xff0c;\u5b83\u4eec\u4ea7\u751f\u7684\u8f68\u8ff9\u5df2\u5927\u89c4\u6a21\u79ef\u7d2f&#xff0c;\u800c\u771f\u5b9e\u7684\u3001\u53ef\u6267\u884c\u7684\u73af\u5883\u4ecd\u7136\u7a00\u7f3a\u3002\u8fd9\u79cd\u5dee\u8ddd\u5f88\u91cd\u8981&#xff0c;\u56e0\u4e3a\u8f68\u8ff9\u548c\u73af\u5883\u7684\u4ef7\u503c\u5e76\u4e0d\u7b49\u540c\u3002\u4e00\u6761\u8f68\u8ff9\u662f\u4e00\u4e2a\u56fa\u5b9a\u7684\u3001\u5355\u4e00\u7684\u8bb0\u5f55&#xff1a;\u5176\u8d28\u91cf\u53d7\u9650\u4e8e\u4ea7\u751f\u5b83\u7684\u7b56\u7565\u6a21\u578b\u7684\u54cd\u5e94&#xff0c;\u5e76\u4e14\u6211\u4eec\u65e0\u6cd5\u68c0\u67e5\u5b83\u5bf9\u4ee3\u7801\u5e93\u6240\u505a\u7684\u66f4\u6539\u662f\u5426\u6b63\u786e\u3002\u800c\u73af\u5883\u5219\u6ca1\u6709\u8fd9\u4e9b\u9650\u5236&#xff0c;\u56e0\u4e3a\u540c\u4e00\u4e2a\u4efb\u52a1\u53ef\u4ee5\u7531\u66f4\u5f3a\u7684\u6a21\u578b\u91cd\u65b0\u89e3\u51b3&#xff0c;\u7ed3\u679c\u53ef\u4ee5\u7531\u6211\u4eec\u81ea\u5df1\u7684\u6d4b\u8bd5\u6765\u9a8c\u8bc1&#xff0c;\u5e76\u4e14\u53ef\u4ee5\u5728\u540c\u4e00\u4e2a\u5de5\u4f5c\u7a7a\u95f4\u4e0a\u63d0\u51fa\u66f4\u96be\u7684\u4efb\u52a1\u3002\u56e0\u6b64&#xff0c;\u5bf9\u4e8e\u540e\u8bad\u7ec3\u800c\u8a00&#xff0c;\u73af\u5883\u624d\u662f\u503c\u5f97\u6269\u5c55\u7684\u8d44\u6e90\u3002\u4e13\u5bb6\u7f16\u5199\u7684\u57fa\u51c6\u6d4b\u8bd5\u5c55\u793a\u4e86\u7406\u60f3\u73af\u5883\u7684\u6837\u5b50\u3002\u4f8b\u5982&#xff0c;Terminal-Bench (Merrill et al., 2026) \u4e3a\u6bcf\u4e2a\u4efb\u52a1\u90fd\u914d\u6709\u4e00\u4e2a\u5b9a\u5236\u5bb9\u5668\u3001\u4e00\u4e2a\u6307\u4ee4\u548c\u4e00\u4e2a\u53ef\u6267\u884c\u7684\u9a8c\u8bc1\u5668\u3002\u8fd9\u79cd\u53ef\u9760\u6027\u6765\u6e90\u4e8e\u4eba\u5de5\u52aa\u529b&#xff0c;\u8fd9\u4e5f\u9650\u5236\u4e86\u53ef\u4ee5\u6784\u5efa\u7684\u4efb\u52a1\u6570\u91cf&#xff0c;\u800c\u6709\u6548\u7684\u540e\u8bad\u7ec3\u9700\u8981\u6bd4\u4eba\u5de5\u7b56\u5212\u591a\u5f97\u591a\u7684\u73af\u5883\u3002\u56e0\u6b64&#xff0c;\u63d0\u4f9b\u5927\u89c4\u6a21\u3001\u771f\u5b9e\u3001\u53ef\u9a8c\u8bc1\u7684\u8bad\u7ec3\u73af\u5883\u662f\u6211\u4eec\u9762\u4e34\u7684\u6311\u6218\u3002<\/p>\n<p>\u73b0\u6709\u5de5\u4f5c\u901a\u8fc7\u4e09\u79cd\u65b9\u5f0f\u6269\u5c55\u53ef\u6267\u884c\u73af\u5883\u3002\u57fa\u4e8e\u4ee3\u7801\u4ed3\u5e93&#xff08;Repository&#xff09;\u7684\u65b9\u6cd5\u4ece\u771f\u5b9e\u4ee3\u7801\u4ed3\u5e93\u7684 git \u5386\u53f2\u4e2d\u6784\u5efa\u4efb\u52a1 (Pan et al., 2024; Jain et al., 2025)\u3002\u5bf9\u4e8e\u8fc7\u53bb\u4fee\u590d\u7684\u4e00\u4e2a\u9519\u8bef&#xff0c;\u4ed6\u4eec\u4f7f\u7528 git \u5c06\u4ee3\u7801\u4ed3\u5e93\u56de\u6eda\u5230\u4fee\u590d\u524d\u7684\u72b6\u6001\u4ee5\u5f62\u6210\u73af\u5883&#xff0c;\u5c06\u539f\u59cb\u9519\u8bef\u62a5\u544a\u4f5c\u4e3a\u4efb\u52a1&#xff0c;\u5e76\u590d\u7528\u968f\u4fee\u590d\u800c\u6765\u7684\u6d4b\u8bd5\u6765\u68c0\u67e5\u89e3\u51b3\u65b9\u6848\u3002\u6270\u52a8&#xff08;Perturbation&#xff09;\u65b9\u6cd5\u9009\u53d6\u4e00\u4e2a\u901a\u8fc7\u5176\u6d4b\u8bd5\u7684\u4ee3\u7801\u4ed3\u5e93&#xff0c;\u6ce8\u5165\u4e00\u4e2a\u9519\u8bef\u4f7f\u5f97\u67d0\u4e9b\u6d4b\u8bd5\u73b0\u5728\u5931\u8d25&#xff0c;\u5e76\u8981\u6c42\u667a\u80fd\u4f53\u4fee\u590d\u5b83 (Yang et al., 2025; Lin et al., 2026)\u3002\u5c11\u6570\u51e0\u4e2a\u4ee3\u7801\u4ed3\u5e93\u53ef\u4ee5\u901a\u8fc7\u8fd9\u79cd\u65b9\u5f0f\u4ea7\u751f\u8bb8\u591a\u4efb\u52a1&#xff0c;\u4f46\u6bcf\u4e2a\u4efb\u52a1\u90fd\u662f\u4fee\u590d\u4efb\u52a1&#xff0c;\u4e14\u4efb\u52a1\u8303\u56f4\u53d7\u9650\u4e8e\u6e90\u4ee3\u7801\u4ed3\u5e93\u548c\u53ef\u6ce8\u5165\u7684\u9519\u8bef\u7c7b\u578b\u3002\u4efb\u52a1\u6761\u4ef6\u5408\u6210&#xff08;Task-conditioned synthesis&#xff09;\u6839\u636e\u7c7b\u522b\u6216\u7ec4\u5408\u8f74 (Gandhi et al., 2026; Ivison et al., 2026)\u3001\u6280\u80fd\u5206\u7c7b\u6cd5 (Hua et al., 2026) \u6216\u6280\u80fd\u56fe (Fan et al., 2026) \u7684\u6307\u5bfc&#xff0c;\u4ece\u5934\u5f00\u59cb\u4e00\u8d77\u751f\u6210\u4efb\u52a1\u53ca\u5176\u73af\u5883\u3002\u8fd9\u4e9b\u65b9\u6cd5\u63d0\u4f9b\u4e86\u5bf9\u8986\u76d6\u8303\u56f4\u548c\u4efb\u52a1-\u73af\u5883\u5bf9\u9f50\u7684\u63a7\u5236&#xff0c;\u4f46\u73af\u5883\u4e0d\u4e0e\u4efb\u4f55\u771f\u5b9e\u9879\u76ee\u5173\u8054&#xff0c;\u56e0\u6b64\u5176\u771f\u5b9e\u6027\u5b8c\u5168\u4f9d\u8d56\u4e8e\u751f\u6210\u5668&#xff0c;\u800c\u751f\u6210\u5668\u503e\u5411\u4e8e\u4ea7\u751f\u5c0f\u578b\u3001\u6574\u6d01\u7684\u5de5\u4f5c\u7a7a\u95f4&#xff0c;\u800c\u975e\u771f\u5b9e\u4ee3\u7801\u3002<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" alt=\"\" height=\"1256\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260905232625-6a9ca521457b2.png\" width=\"2432\" \/><\/p>\n<p>\u56fe 1&#xff1a;Terminal-Universe \u6982\u89c8\u3002\u8f68\u8ff9\u548c\u73af\u5883\u662f\u540c\u4e00 episode \u7684\u4e24\u4e2a\u89c6\u56fe&#xff0c;\u56e0\u6b64\u5148\u524d\u7684\u5de5\u4f5c\u4ece\u73af\u5883\u4e2d\u5c55\u5f00\u8f68\u8ff9&#xff0c;\u800c\u6211\u4eec\u5219\u53cd\u8f6c\u6620\u5c04&#xff0c;\u4ece\u8f68\u8ff9\u4e2d\u6062\u590d\u73af\u5883&#xff08;\u5de6&#xff09;\u3002\u91cd\u5efa\u5206\u4e24\u4e2a\u9636\u6bb5\u8fdb\u884c&#xff0c;\u5148\u662f\u786e\u5b9a\u6027\u56de\u653e&#xff0c;\u7136\u540e\u662f\u667a\u80fd\u4f53\u8865\u5168&#xff08;\u4e2d&#xff09;\u3002\u6bcf\u4e2a\u6062\u590d\u7684\u73af\u5883\u968f\u540e\u5728\u5355\u4e2a\u5de5\u4f5c\u7a7a\u95f4\u5185\u3001\u8de8\u591a\u4e2a\u4f9d\u8d56\u5de5\u4f5c\u7a7a\u95f4\u6216\u591a\u4e2a\u7528\u6237\u8f6e\u6b21\u4e2d\u88ab\u91cd\u65b0\u67e5\u8be2&#xff08;\u53f3&#xff09;\u3002\u56fe 2 \u8be6\u7ec6\u8bf4\u660e\u4e86\u6bcf\u79cd\u673a\u5236\u3002<\/p>\n<p>\u5728\u8fd9\u4e9b\u9014\u5f84\u4e2d&#xff0c;\u73af\u5883\u6784\u5efa\u8981\u4e48\u4ece\u73b0\u6709\u73af\u5883\u5f00\u59cb&#xff0c;\u8981\u4e48\u4e0e\u4ece\u5934\u751f\u6210\u7684\u65b0\u4efb\u52a1\u8026\u5408\u3002\u8f68\u8ff9&#xff0c;\u4f5c\u4e3a\u5b83\u4eec\u8fd0\u884c\u6240\u5728\u73af\u5883\u7684\u89c2\u6d4b&#xff0c;\u4f5c\u4e3a\u4e00\u79cd\u8d44\u6e90\u5728\u5f88\u5927\u7a0b\u5ea6\u4e0a\u4ecd\u672a\u88ab\u5145\u5206\u63a2\u7d22\u3002\u8f68\u8ff9\u4e2d\u8bb0\u5f55\u7684\u5de5\u5177\u8c03\u7528\u5df2\u7ecf\u63ed\u793a\u4e86\u5176\u8fd0\u884c\u73af\u5883\u7684\u5185\u5bb9\u548c\u7ed3\u6784\u3002\u4f8b\u5982&#xff0c;\u5de5\u5177 Read \u663e\u793a\u6587\u4ef6\u5185\u5bb9&#xff0c;Write \u548c Edit \u663e\u793a\u5de5\u4f5c\u7a7a\u95f4\u5982\u4f55\u53d8\u5316\u3002\u8fd9\u8db3\u4ee5\u91cd\u5efa\u4e00\u4e2a\u53ef\u6267\u884c\u7684\u539f\u59cb\u5de5\u4f5c\u7a7a\u95f4\u526f\u672c\u3002\u56e0\u6b64&#xff0c;\u6211\u4eec\u63d0\u51fa\u4e86 Terminal-Universe (\u56fe 1)&#xff0c;\u4e00\u4e2a\u6846\u67b6&#xff0c;\u5b83\u5c06\u6bcf\u6761\u8f68\u8ff9\u8f6c\u5316\u4e3a\u4e00\u4e2a\u53ef\u91cd\u7528\u7684\u73af\u5883&#xff0c;\u5e76\u5bf9\u5176\u8fdb\u884c\u63a2\u7d22\u4ee5\u5408\u6210\u65b0\u4efb\u52a1\u548c\u8fdb\u884c\u6301\u7eed\u4ea4\u4e92\u3002<\/p>\n<p>\u5177\u4f53\u6765\u8bf4&#xff0c;\u4e3a\u4e86\u57fa\u4e8e\u8f68\u8ff9\u6784\u5efa\u73af\u5883 (\u00a73.1)&#xff0c;\u6211\u4eec\u9996\u5148\u56de\u653e\u8bb0\u5f55\u7684\u6587\u4ef6\u64cd\u4f5c&#xff0c;\u5e76\u5c06\u6bcf\u4e2a\u6587\u4ef6\u6062\u590d\u5230\u667a\u80fd\u4f53\u66f4\u6539\u4e4b\u524d\u7684\u72b6\u6001&#xff0c;\u4fdd\u7559\u667a\u80fd\u4f53\u81ea\u8eab\u7684\u7f16\u8f91&#xff0c;\u4ee5\u4fbf\u5de5\u4f5c\u7a7a\u95f4\u4ee5\u672a\u89e3\u51b3\u7684\u72b6\u6001\u5f00\u59cb\u3002\u7531\u4e8e\u56de\u653e\u901a\u5e38\u53ea\u4ea7\u751f\u90e8\u5206\u5de5\u4f5c\u7a7a\u95f4&#xff0c;\u4e00\u4e2a\u8865\u5168\u667a\u80fd\u4f53\u4f1a\u81ea\u52a8\u63d0\u4f9b\u4efb\u52a1\u6240\u9700\u7684\u7f3a\u5931\u6587\u4ef6\u548c\u4f9d\u8d56\u9879&#xff0c;\u540c\u65f6\u4e0d\u6cc4\u9732\u89e3\u51b3\u65b9\u6848\u3002\u5bf9\u4e8e\u76f8\u5e94\u7684\u4efb\u52a1&#xff0c;\u6211\u4eec\u65e2\u91cd\u5efa\u539f\u59cb\u610f\u56fe\u67e5\u8be2&#xff0c;\u4e5f\u5408\u6210\u5168\u65b0\u7684\u67e5\u8be2\u3002\u6b64\u5916&#xff0c;\u4e3a\u4e86\u66f4\u597d\u5730\u5339\u914d\u771f\u5b9e\u4e16\u754c\u7684\u8f6f\u4ef6\u5de5\u7a0b\u573a\u666f&#xff0c;\u6211\u4eec\u8fd8\u6cbf\u7740\u4e24\u4e2a\u4e92\u8865\u7684\u7ef4\u5ea6\u6269\u5c55\u73af\u5883\u4e0a\u7684\u4efb\u52a1&#xff1a;\u5e7f\u5ea6\u548c\u6df1\u5ea6 (\u00a73.2)\u3002\u5e7f\u5ea6\u6269\u5c55\u8d85\u8d8a\u4e86\u5927\u591a\u6570\u4efb\u52a1\u5408\u6210\u6240\u5047\u8bbe\u7684\u5355\u4e00\u4ee3\u7801\u4ed3\u5e93\u8bbe\u7f6e\u3002\u6211\u4eec\u8bc6\u522b\u76f8\u5173\u73af\u5883\u4e4b\u95f4\u7684\u4f9d\u8d56\u5173\u7cfb&#xff0c;\u5e76\u6784\u5efa\u8de8\u8d8a\u591a\u4e2a\u4ee3\u7801\u5e93\u7684\u8de8\u5de5\u4f5c\u7a7a\u95f4\u4efb\u52a1&#xff0c;\u8fd9\u4e9b\u4efb\u52a1\u96be\u5ea6\u66f4\u5927&#xff0c;\u4f8b\u5982&#xff0c;\u9605\u8bfb\u53c2\u8003\u5b9e\u73b0\u3001\u5c06\u67d0\u4e2a\u7279\u6027\u4ece\u4e00\u4e2a\u9879\u76ee\u8fc1\u79fb\u5230\u53e6\u4e00\u4e2a\u9879\u76ee&#xff0c;\u6216\u8fde\u63a5\u4e24\u4e2a\u7ec4\u4ef6\u3002\u6df1\u5ea6\u6269\u5c55\u5c06\u5355\u8f6e\u4efb\u52a1\u8f6c\u53d8\u4e3a\u591a\u8f6e\u4efb\u52a1\u3002\u667a\u80fd\u4f53\u5b8c\u6210\u521d\u59cb\u67e5\u8be2\u540e&#xff0c;\u4e00\u4e2a\u7528\u6237\u667a\u80fd\u4f53\u4f1a\u968f\u7740\u5de5\u4f5c\u7a7a\u95f4\u7684\u53d8\u5316\u63d0\u51fa\u6709\u6839\u636e\u7684\u540e\u7eed\u95ee\u9898\u2014\u2014\u7528\u65b0\u9700\u6c42\u6269\u5c55\u5b83&#xff0c;\u6216\u8005&#xff0c;\u5f53\u4e00\u8f6e\u5931\u8d25\u65f6&#xff0c;\u6839\u636e\u51fa\u9519\u539f\u56e0\u8981\u6c42\u4fee\u590d\u3002\u6211\u4eec\u9a8c\u8bc1\u6bcf\u4e00\u8f6e&#xff0c;\u5e76\u5c06\u7ed3\u679c\u4f5c\u4e3a\u81ea\u7136\u7684\u3001\u7528\u6237\u53ef\u89c1\u7684\u53cd\u9988\u53cd\u9988\u7ed9\u667a\u80fd\u4f53\u3002\u6bcf\u4e2a\u65b0\u4efb\u52a1\u90fd\u9644\u5e26\u4e00\u4e2a\u7531\u667a\u80fd\u4f53\u5728\u5bb9\u5668\u5185\u7f16\u5199\u7684\u9a8c\u8bc1\u5668&#xff0c;\u6211\u4eec\u53ea\u4fdd\u7559\u6240\u6709\u6d4b\u8bd5\u90fd\u901a\u8fc7\u7684\u8f68\u8ff9 (\u00a74)\u3002<\/p>\n<p>\u5728\u516c\u5f00\u53ef\u7528\u7684\u7ec8\u7aef\u667a\u80fd\u4f53\u8f68\u8ff9\u4e0a&#xff0c;\u8be5\u6d41\u6c34\u7ebf\u4ea7\u751f\u4e86 37.3k \u4e2a\u4efb\u52a1\u5145\u5206\u7684\u73af\u5883\u3002\u5728\u8fd9\u4e9b\u73af\u5883\u4e4b\u4e0a&#xff0c;\u6211\u4eec\u751f\u6210\u65b0\u7684\u67e5\u8be2&#xff0c;\u5e76\u4f7f\u7528 Qwen3.7-Max \u4f5c\u4e3a\u6559\u5e08\u4e3a\u6bcf\u4e2a\u4efb\u52a1\u5c55\u5f00\u4e00\u4e2a\u89e3\u51b3\u65b9\u6848\u8f68\u8ff9\u3002\u5728\u751f\u6210\u7684\u6570\u636e\u4e0a\u8bad\u7ec3 Qwen3.5-27B \u9a8c\u8bc1\u4e86\u6211\u4eec\u65b9\u6cd5\u7684\u6709\u6548\u6027&#xff1a;\u5b83\u5728\u5355\u8f6e\u57fa\u51c6\u6d4b\u8bd5 Terminal-Bench 2.1 \u4e0a\u63d0\u9ad8\u4e86 11.9 \u4e2a\u70b9&#xff0c;\u5728\u591a\u8f6e\u57fa\u51c6\u6d4b\u8bd5 EvoCode-Bench v2 MT&#064;4 \u4e0a\u63d0\u9ad8\u4e86 13.8 \u4e2a\u70b9 (\u00a75)\u3002\u5927\u91cf\u7684\u6d88\u878d\u5b9e\u9a8c\u8bc1\u5b9e\u6bcf\u4e2a\u7ec4\u4ef6\u90fd\u6709\u8d21\u732e&#xff0c;\u6700\u503c\u5f97\u6ce8\u610f\u7684\u662f&#xff0c;\u5728\u91cd\u5efa\u73af\u5883\u4e2d\u91cd\u65b0\u89e3\u51b3\u95ee\u9898\u8fdc\u8fdc\u4f18\u4e8e\u6a21\u4eff\u539f\u59cb\u8f68\u8ff9\u3002<\/p>\n<p>\u8868 1&#xff1a;\u4ee3\u8868\u6027\u73af\u5883\u548c\u8bad\u7ec3\u6570\u636e\u6784\u5efa\u65b9\u6cd5\u7684\u6bd4\u8f83\u3002\u7b56\u7565\u603b\u7ed3\u4e86\u7528\u4e8e\u6784\u5efa\u73af\u5883\u6216\u8bad\u7ec3\u4efb\u52a1\u7684\u4e3b\u8981\u673a\u5236\u3002\u73af\u5883\u8ba1\u6570\u7edf\u8ba1\u53ef\u91cd\u7528\u7684\u6e90\u4ee3\u7801\u4ed3\u5e93\/\u5de5\u4f5c\u7a7a\u95f4\u6216\u72ec\u7acb\u6784\u5efa\u7684\u73af\u5883&#xff0c;\u4efb\u52a1\u6570\u7ed9\u51fa\u62a5\u544a\u7684\u4efb\u52a1\u6216\u8bad\u7ec3\u5b9e\u4f8b\u6570\u91cf\u3002\u9a8c\u8bc1\u8868\u793a\u53ef\u6267\u884c\u7684\u4efb\u52a1\u9a8c\u8bc1\u5668&#xff0c;\u591a\u8f6e\u8868\u793a\u6269\u5c55\u7684\u591a\u8f6e\u67e5\u8be2&#xff0c;\u8de8\u5de5\u4f5c\u7a7a\u95f4\u8868\u793a\u8de8\u5de5\u4f5c\u7a7a\u95f4\u5408\u6210\u3002<\/p>\n<table>\n<tr>\u65b9\u6cd5\u7b56\u7565\u73af\u5883\u4efb\u52a1\u9a8c\u8bc1\u591a\u8f6e\u8de8\u5de5\u4f5c\u7a7a\u95f4<\/tr>\n<tbody>\n<tr>\n<td>Endless Term. (Gandhi et al., 2026)<\/td>\n<td>\u4efb\u52a1\u6761\u4ef6<\/td>\n<td>3,255<\/td>\n<td>3,255<\/td>\n<td>\u2713<\/td>\n<td>\u00d7<\/td>\n<td>\u00d7<\/td>\n<\/tr>\n<tr>\n<td>TMax (Ivison et al., 2026)<\/td>\n<td>\u7ec4\u5408\u91c7\u6837<\/td>\n<td>14.6k<\/td>\n<td>14.6k<\/td>\n<td>\u2713<\/td>\n<td>\u00d7<\/td>\n<td>\u00d7<\/td>\n<\/tr>\n<tr>\n<td>CLI-Gym (Lin et al., 2026)<\/td>\n<td>\u73af\u5883\u6270\u52a8<\/td>\n<td>291,655<\/td>\n<td>\u2713<\/td>\n<td>\u00d7<\/td>\n<td>\u00d7<\/td>\n<td>\u00d7<\/td>\n<\/tr>\n<tr>\n<td>CLI-Universe (Hua et al., 2026)<\/td>\n<td>\u5206\u7c7b\u6cd5\u5f15\u5bfc<\/td>\n<td>6k<\/td>\n<td>6k<\/td>\n<td>\u2713<\/td>\n<td>\u00d7<\/td>\n<td>\u00d7<\/td>\n<\/tr>\n<tr>\n<td>SkillSynth (Fan et al., 2026)<\/td>\n<td>\u6280\u80fd\u56fe\u5f15\u5bfc<\/td>\n<td>3,560<\/td>\n<td>3,560<\/td>\n<td>\u2713<\/td>\n<td>\u00d7<\/td>\n<td>\u00d7<\/td>\n<\/tr>\n<tr>\n<td>OpenThinker-Agent (Raof et al., 2026)<\/td>\n<td>\u591a\u6e90\u7b56\u5c55<\/td>\n<td>&#8211;<\/td>\n<td>100k<\/td>\n<td>\u2713<\/td>\n<td>\u00d7<\/td>\n<td>\u00d7<\/td>\n<\/tr>\n<tr>\n<td>RST (Li et al., 2026b)<\/td>\n<td>\u9012\u5f52\u6f14\u5316<\/td>\n<td>37.5k<\/td>\n<td>37.5k<\/td>\n<td>\u2713<\/td>\n<td>\u00d7<\/td>\n<td>\u00d7<\/td>\n<\/tr>\n<tr>\n<td>CalibForge (Meng et al., 2026)<\/td>\n<td>\u5bf9\u6297\u6027\u6821\u51c6<\/td>\n<td>5,431<\/td>\n<td>5,431<\/td>\n<td>\u2713<\/td>\n<td>\u00d7<\/td>\n<td>\u00d7<\/td>\n<\/tr>\n<tr>\n<td>Terminal-Universe (ours)<\/td>\n<td>\u8f68\u8ff9\u91cd\u5efa<\/td>\n<td>37.3k<\/td>\n<td>32.0k<\/td>\n<td>\u2713<\/td>\n<td>\u2713<\/td>\n<td>\u2713<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u603b\u4e4b&#xff0c;\u6211\u4eec\u505a\u51fa\u4ee5\u4e0b\u8d21\u732e&#xff1a;<\/p>\n<ul>\n<li>\n<p>\u73af\u5883\u91cd\u5efa\u3002\u6211\u4eec\u5c06\u8bb0\u5f55\u7684\u667a\u80fd\u4f53\u8f68\u8ff9\u91cd\u65b0\u5b9a\u4e49\u4e3a\u53ef\u91cd\u7528\u53ef\u6267\u884c\u73af\u5883\u7684\u6765\u6e90&#xff0c;\u5e76\u901a\u8fc7\u786e\u5b9a\u6027\u56de\u653e\u548c\u667a\u80fd\u4f53\u8865\u5168\u6765\u91cd\u5efa\u6bcf\u4e00\u4e2a\u2014\u2014\u65e0\u9700\u539f\u59cb\u4ee3\u7801\u4ed3\u5e93\u6216\u4ece\u5934\u6784\u5efa\u3002<\/p>\n<\/li>\n<li>\n<p>\u91cd\u65b0\u67e5\u8be2\u65b9\u6cd5\u3002\u6211\u4eec\u901a\u8fc7\u751f\u6210\u65b0\u4efb\u52a1\u6765\u6269\u5c55\u6bcf\u4e2a\u91cd\u5efa\u73af\u5883\u7684\u6548\u7528\u3002\u9664\u4e86\u6807\u51c6\u7684\u5355\u5de5\u4f5c\u7a7a\u95f4\u3001\u5355\u8f6e\u4efb\u52a1\u5916&#xff0c;\u6211\u4eec\u8fd8\u901a\u8fc7\u8de8\u5de5\u4f5c\u7a7a\u95f4\u4efb\u52a1\u5408\u6210\u8d21\u732e\u4e86\u5e7f\u5ea6\u6269\u5c55&#xff0c;\u5e76\u901a\u8fc7\u591a\u8f6e\u7528\u6237\u67e5\u8be2\u8d21\u732e\u4e86\u6df1\u5ea6\u6269\u5c55\u3002\u6bcf\u4e2a\u4efb\u52a1\u90fd\u914d\u6709\u4e00\u4e2a\u667a\u80fd\u4f53\u64b0\u5199\u7684\u9a8c\u8bc1\u5668&#xff0c;\u53ea\u6709\u901a\u8fc7\u7684\u8f68\u8ff9\u624d\u88ab\u4fdd\u7559\u3002<\/p>\n<\/li>\n<li>\n<p>\u5b9e\u8bc1\u9a8c\u8bc1\u3002\u5728\u751f\u6210\u7684\u8bed\u6599\u5e93\u4e0a\u5fae\u8c03 Qwen3.5-27B&#xff0c;\u5728 Terminal-Bench 2.1 \u4e0a\u6bd4\u540c\u4e00\u57fa\u7840\u6a21\u578b\u63d0\u9ad8\u4e86 11.9 \u4e2a\u70b9&#xff0c;\u5728 EvoCode-Bench v2 MT&#064;4 \u4e0a\u63d0\u9ad8\u4e86 13.8 \u4e2a\u70b9\u3002\u6d88\u878d\u5b9e\u9a8c\u8bc1\u5b9e\u6bcf\u4e2a\u7ec4\u4ef6\u90fd\u6709\u8d21\u732e&#xff0c;\u5c24\u5176\u503c\u5f97\u6ce8\u610f\u7684\u662f&#xff0c;\u5728\u91cd\u5efa\u73af\u5883\u4e2d\u91cd\u65b0\u89e3\u51b3\u95ee\u9898\u8fdc\u8fdc\u4f18\u4e8e\u6a21\u4eff\u539f\u59cb\u8f68\u8ff9\u3002<\/p>\n<\/li>\n<\/ul>\n<h3>2 \u76f8\u5173\u5de5\u4f5c<\/h3>\n<p>\u7ec8\u7aef\u73af\u5883\u6269\u5c55\u3002\u00a0\u73b0\u6709\u65b9\u6cd5\u901a\u8fc7\u4e09\u79cd\u4e3b\u8981\u9014\u5f84\u6269\u5c55\u53ef\u6267\u884c\u5de5\u4f5c\u7a7a\u95f4&#xff1a;\u4ece\u771f\u5b9e\u5f00\u53d1\u5386\u53f2\u4e2d\u6062\u590d\u4ee3\u7801\u4ed3\u5e93\u72b6\u6001\u3001\u4fee\u6539\u6b63\u5e38\u8fd0\u884c\u7684\u5de5\u4f5c\u7a7a\u95f4&#xff0c;\u6216\u4ece\u751f\u6210\u7684\u89c4\u683c\u8bf4\u660e\u6784\u5efa\u7279\u5b9a\u4efb\u52a1\u7684\u73af\u5883\u3002SWE-Gym (Pan et al., 2024) \u548c R2E-Gym (Jain et al., 2025) \u4ece\u4e0e\u5386\u53f2\u95ee\u9898\u6216\u63d0\u4ea4\u76f8\u5173\u7684\u4ee3\u7801\u4ed3\u5e93\u7248\u672c\u6784\u5efa\u73af\u5883\u3002SWE-smith (Yang et al., 2025) \u548c CLI-Gym (Lin et al., 2026) \u4ece\u6b63\u5e38\u8fd0\u884c\u7684\u4ee3\u7801\u4ed3\u5e93\u6216 CLI \u5de5\u4f5c\u7a7a\u95f4\u5f00\u59cb&#xff0c;\u5e76\u5411\u5176\u4e2d\u5f15\u5165\u9519\u8bef\u6216\u6545\u969c\u3002SETA (Shen et al., 2026b) \u5219\u901a\u8fc7\u5408\u6210\u548c\u81ea\u9002\u5e94\u73af\u5883\u6f14\u5316\u6765\u6269\u5c55\u53ef\u9a8c\u8bc1\u7684\u7ec8\u7aef\u73af\u5883\u4ee5\u7528\u4e8e\u5f3a\u5316\u5b66\u4e60\u3002\u53e6\u4e00\u4e2a\u72ec\u7acb\u7684\u5bb6\u65cf\u8054\u5408\u751f\u6210\u7ec8\u7aef\u4efb\u52a1\u53ca\u5176\u5bb9\u5668\u5316\u7684\u6267\u884c\u73af\u5883&#xff1b;\u6211\u4eec\u4e0b\u9762\u8ba8\u8bba\u5b83\u4eec\u7684\u4efb\u52a1\u6784\u5efa\u7b56\u7565\u3002Terminal-Universe \u5219\u5c06\u8bb0\u5f55\u7684\u667a\u80fd\u4f53\u8f68\u8ff9\u4f5c\u4e3a\u8d77\u70b9\u3002\u5b83\u56de\u653e\u6bcf\u6761\u8f68\u8ff9\u4e2d\u8bb0\u5f55\u7684\u6587\u4ef6\u64cd\u4f5c&#xff0c;\u5e76\u4f7f\u7528\u667a\u80fd\u4f53\u8865\u5168\u6765\u6062\u590d\u7f3a\u5931\u7684\u9879\u76ee\u4e0a\u4e0b\u6587&#xff0c;\u4ea7\u751f\u53ef\u91cd\u7528\u3001\u53ef\u6267\u884c\u7684\u5de5\u4f5c\u7a7a\u95f4\u3002<\/p>\n<p>\u7ec8\u7aef\u4efb\u52a1\u5408\u6210\u3002\u00a0\u73b0\u6709\u65b9\u6cd5\u4ece\u62bd\u8c61\u5148\u9a8c\u6216\u5177\u4f53\u53ef\u6267\u884c\u73af\u5883\u4e2d\u63a8\u5bfc\u4efb\u52a1\u3002\u81ea\u9876\u5411\u4e0b&#xff08;Top-down&#xff09;\u6d41\u6c34\u7ebf\u4ece\u7c7b\u522b\u6216\u79cd\u5b50\u751f\u6210\u4efb\u52a1-\u73af\u5883\u5bf9 (Gandhi et al., 2026; Pi et al., 2026; Zhu et al., 2026)&#xff0c;\u800c\u66f4\u591a\u7ed3\u6784\u5316\u53d8\u4f53\u56f4\u7ed5\u80fd\u529b\u5206\u7c7b\u3001\u7ec4\u5408\u8f74\u6216\u6280\u80fd\u56fe\u7ec4\u7ec7\u751f\u6210\u8fc7\u7a0b (Hua et al., 2026; Ivison et al., 2026; Fan et al., 2026)\u3002\u5176\u4ed6\u65b9\u6cd5\u4f7f\u7528\u667a\u80fd\u4f53\u6280\u80fd\u3001\u53ef\u6267\u884c\u5143\u4efb\u52a1\u6216\u6c42\u89e3\u5668\u53cd\u9988\u6765\u591a\u6837\u5316\u548c\u6821\u51c6\u751f\u6210\u7684\u4efb\u52a1 (Cheng et al., 2026; Pan et al., 2026; Meng et al., 2026)\u3002<\/p>\n<p>\u57fa\u4e8e\u73af\u5883&#xff08;Environment-grounded&#xff09;\u7684\u65b9\u6cd5\u5219\u4ece\u5de5\u4f5c\u4ee3\u7801\u548c\u9879\u76ee\u4e0a\u4e0b\u6587\u4e2d\u63a8\u5bfc\u4efb\u52a1\u3002\u5b83\u4eec\u6270\u52a8\u5065\u5eb7\u73af\u5883\u3001\u6316\u6398\u4ee3\u7801\u4ed3\u5e93\u6587\u6863\u3001\u5c06\u4efb\u52a1\u57fa\u4e8e\u771f\u5b9e\u95ee\u9898&#xff0c;\u6216\u5728\u5171\u4eab\u73af\u5883\u4e2d\u8054\u5408\u5b9e\u73b0\u6307\u4ee4\u3001\u89e3\u51b3\u65b9\u6848\u548c\u9a8c\u8bc1\u5668 (Lin et al., 2026; Wu et al., 2026a; Yang et al., 2026; Shi et al., 2026)\u3002RST \u901a\u8fc7\u5ef6\u957f\u89e3\u51b3\u65b9\u6848\u5e76\u91cd\u65b0\u5bf9\u9f50\u5176\u4efb\u52a1\u3001\u9a8c\u8bc1\u5668\u548c\u73af\u5883\u6765\u9012\u5f52\u6269\u5c55\u5df2\u9a8c\u8bc1\u7684\u79cd\u5b50 (Li et al., 2026b)\u3002Terminal-Universe \u5c06\u8fc7\u53bb\u7684\u667a\u80fd\u4f53\u8f68\u8ff9\u4f5c\u4e3a\u8d77\u70b9&#xff1a;\u5b83\u5728\u751f\u6210\u65b0\u4efb\u52a1\u4e4b\u524d\u91cd\u5efa\u5b83\u4eec\u7684\u5de5\u4f5c\u7a7a\u95f4&#xff0c;\u65e0\u8bba\u662f\u5728\u5de5\u4f5c\u7a7a\u95f4\u5185\u90e8\u8fd8\u662f\u8de8\u5de5\u4f5c\u7a7a\u95f4\u3002<\/p>\n<p>\u4ea4\u4e92\u5f0f\u548c\u591a\u8f6e\u667a\u80fd\u4f53\u3002\u00a0\u8d85\u8d8a\u5b64\u7acb\u7684\u3001\u5355\u8f6e\u63d0\u793a&#xff0c;\u6700\u8fd1\u7684\u57fa\u51c6\u6d4b\u8bd5\u5c06\u667a\u80fd\u4f53\u4ea4\u4e92\u5efa\u6a21\u4e3a\u52a8\u6001\u7684\u591a\u8f6e\u5bf9\u8bdd\u3002InterCode \u901a\u8fc7\u5bb9\u5668\u5316 shell \u4e2d\u7684\u6267\u884c\u53cd\u9988\u5f62\u5f0f\u5316\u4ea4\u4e92\u5f0f\u7f16\u7801 (Yang et al., 2023)\u3002SWE-INTERACT (Raghavendra et al., 2026) \u4f7f\u7528\u6a21\u62df\u7528\u6237\u9010\u6b65\u63ed\u793a\u9700\u6c42\u5e76\u63d0\u4f9b\u6709\u9488\u5bf9\u6027\u7684\u4fee\u8ba2&#xff0c;\u800c SWE-Together (Wu et al., 2026b) \u901a\u8fc7\u4e00\u4e2a\u72b6\u6001\u6761\u4ef6\u7528\u6237\u6a21\u62df\u5668\u56de\u653e\u771f\u5b9e\u7684\u7528\u6237-\u667a\u80fd\u4f53\u4f1a\u8bdd&#xff0c;\u5728\u88ab\u8bc4\u4f30\u7684\u667a\u80fd\u4f53\u53d6\u5f97\u8fdb\u5c55\u65f6\u63d0\u4f9b\u53cd\u9988\u3002ICAE-Bench (Peng et al., 2026b) \u4f7f\u7528\u81ea\u52a8\u5316\u7528\u6237\u667a\u80fd\u4f53\u4ece\u4e0d\u5b8c\u6574\u7684\u4ea7\u54c1\u9700\u6c42\u4e2d\u8bc4\u4f30\u4ea4\u4e92\u5f0f\u9879\u76ee\u6784\u5efa\u3002\u5728\u6301\u4e45\u5de5\u4f5c\u7a7a\u95f4\u4e2d&#xff0c;EvoCode-Bench v2 (Shen et al., 2026a) \u8bc4\u4f30\u667a\u80fd\u4f53\u5728\u8fde\u7eed\u5f00\u53d1\u8bf7\u6c42\u4e2d\u7684\u8868\u73b0\u3002Terminal-Universe \u901a\u8fc7\u6df1\u5ea6\u6269\u5c55\u62e5\u62b1\u8fd9\u79cd\u4ea4\u4e92\u5f0f\u8bbe\u7f6e&#xff0c;\u5c06\u91cd\u5efa\u7684\u5de5\u4f5c\u7a7a\u95f4\u6269\u5c55\u5230\u591a\u8f6e\u4f1a\u8bdd\u4e2d&#xff0c;\u4ee5\u6355\u6349\u8fed\u4ee3\u7684\u7528\u6237\u53cd\u9988\u548c\u9700\u6c42\u7ec6\u5316\u3002<\/p>\n<p>\u8868 1 \u5c06 Terminal-Universe \u4e0e\u4ee3\u8868\u6027\u65b9\u6cd5\u8fdb\u884c\u4e86\u5b9a\u4f4d\u3002\u4e0e\u5148\u524d\u5de5\u4f5c\u76f8\u6bd4&#xff0c;\u5b83\u4ece\u8bb0\u5f55\u7684\u8f68\u8ff9\u6784\u5efa\u73af\u5883&#xff0c;\u5e76\u6cbf\u5e7f\u5ea6&#xff08;\u8de8\u5de5\u4f5c\u7a7a\u95f4\u5408\u6210&#xff09;\u548c\u6df1\u5ea6&#xff08;\u591a\u8f6e\u67e5\u8be2&#xff09;\u4e24\u4e2a\u7ef4\u5ea6\u6269\u5c55\u4efb\u52a1\u3002<\/p>\n<h3>3 Terminal-Universe<\/h3>\n<p>\u56fe 2 \u8be6\u7ec6\u8bf4\u660e\u4e86\u73af\u5883\u91cd\u5efa\u3001\u56db\u79cd\u91cd\u65b0\u67e5\u8be2\u53d8\u4f53\u4ee5\u53ca\u9a8c\u8bc1\u7684\u673a\u5236\u3002\u4ee5\u4e0b\u5c0f\u8282\u5206\u522b\u63cf\u8ff0\u73af\u5883\u91cd\u5efa (\u00a73.1)\u3001\u91cd\u65b0\u67e5\u8be2 (\u00a73.2) \u548c\u9a8c\u8bc1 (\u00a73.3)\u3002<\/p>\n<p class=\"img-center\"><img loading=\"lazy\" decoding=\"async\" alt=\"\" height=\"1162\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260905232626-6a9ca52281aa5.png\" width=\"1368\" \/><\/p>\n<h4>3.1 \u73af\u5883\u91cd\u5efa<\/h4>\n<p>\u4ece\u8f68\u8ff9 \u03c4 \u4e2d\u8bb0\u5f55\u7684\u6587\u4ef6\u548c\u547d\u4ee4\u64cd\u4f5c&#xff0c;\u6211\u4eec\u6062\u590d\u4e00\u4e2a\u53ef\u6267\u884c\u7684\u5de5\u4f5c\u7a7a\u95f4\u00a0E^&#xff0c;\u5b83\u8fd1\u4f3c\u4e8e\u4ea7\u751f \u03c4 \u7684\u73af\u5883\u00a0E\u3002\u8fd9\u79cd\u6062\u590d\u672c\u8d28\u4e0a\u6709\u635f\u7684&#xff0c;\u56e0\u4e3a\u672a\u8bbf\u95ee\u7684\u6587\u4ef6\u3001\u9690\u5f0f\u7cfb\u7edf\u4f9d\u8d56\u9879\u548c\u5916\u90e8\u7f51\u7edc\u8d44\u6e90\u6ca1\u6709\u7559\u4e0b\u76f4\u63a5\u75d5\u8ff9\u3002\u56e0\u6b64&#xff0c;\u6211\u4eec\u5206\u4e09\u4e2a\u9636\u6bb5\u8fdb\u884c\u91cd\u5efa&#xff1a;\u786e\u5b9a\u6027\u56de\u653e\u6062\u590d \u03c4 \u76f4\u63a5\u66b4\u9732\u7684\u6587\u4ef6\u72b6\u6001&#xff0c;\u667a\u80fd\u4f53\u8865\u5168\u63d0\u4f9b\u5176\u9057\u6f0f\u7684\u6f5c\u5728\u4e0a\u4e0b\u6587&#xff0c;\u73af\u5883\u8fc7\u6ee4\u53ea\u4fdd\u7559\u5bf9\u6062\u590d\u7684\u4efb\u52a1 q \u8db3\u591f\u7684\u5de5\u4f5c\u7a7a\u95f4\u3002<\/p>\n<p>\u9636\u6bb5 1&#xff1a;\u786e\u5b9a\u6027\u56de\u653e\u3002\u00a0\u56de\u653e\u6309\u65f6\u95f4\u987a\u5e8f\u5904\u7406 \u03c4 \u4e2d\u7684\u8bfb\u3001\u5199\u548c\u7f16\u8f91\u64cd\u4f5c&#xff0c;\u4ee5\u6062\u590d\u6bcf\u4e2a\u88ab\u8bbf\u95ee\u8def\u5f84\u5728\u8f68\u8ff9\u4e2d\u53ef\u89c1\u7684\u6700\u65e9\u548c\u6700\u65b0\u6587\u4ef6\u5185\u5bb9\u3002\u91cd\u5efa\u7684\u521d\u59cb\u5de5\u4f5c\u7a7a\u95f4 E^0\u200b\u00a0\u6536\u96c6\u6bcf\u4e2a\u9884\u5148\u5b58\u5728\u7684\u6587\u4ef6\u5728\u5176\u6700\u65e9\u89c2\u5bdf\u5230\u7684\u7248\u672c&#xff08;\u5728\u667a\u80fd\u4f53\u9996\u6b21\u66f4\u6539\u4e4b\u524d&#xff09;\u7684\u5185\u5bb9&#xff1b;\u667a\u80fd\u4f53\u521b\u5efa\u7684\u6587\u4ef6\u88ab\u6392\u9664&#xff0c;\u667a\u80fd\u4f53\u7684\u6587\u4ef6\u66f4\u6539\u88ab\u5355\u72ec\u8bb0\u5f55\u4ee5\u4f9b\u540e\u7eed\u9a8c\u8bc1\u3002\u7531\u4e8e\u8f68\u8ff9\u53ea\u66b4\u9732\u667a\u80fd\u4f53\u63a5\u89e6\u8fc7\u7684\u8def\u5f84&#xff0c;\u5e76\u4e14\u5f53\u8f68\u8ff9\u53ea\u663e\u793a\u6587\u4ef6\u7684\u90e8\u5206\u5185\u5bb9\u6216\u7ec8\u7aef\u8f93\u51fa\u88ab\u622a\u65ad\u65f6&#xff0c;\u6587\u4ef6\u5185\u5bb9\u53ef\u80fd\u4e0d\u5b8c\u6574&#xff0c;\u56e0\u6b64\u00a0E^0 \u4ecd\u7136\u662f\u4e00\u4e2a\u90e8\u5206\u5de5\u4f5c\u7a7a\u95f4\u3002<\/p>\n<p>\u9636\u6bb5 2&#xff1a;\u667a\u80fd\u4f53\u8865\u5168\u3002\u00a0\u7ed9\u5b9a\u90e8\u5206\u5de5\u4f5c\u7a7a\u95f4\u00a0E^0\u200b\u00a0\u548c\u6062\u590d\u7684\u4efb\u52a1 q&#xff0c;\u4e00\u4e2a\u8865\u5168\u667a\u80fd\u4f53\u521b\u5efa\u7f3a\u5931\u7684\u6587\u4ef6\u3001\u8865\u5168\u90e8\u5206\u6587\u4ef6&#xff0c;\u5e76\u6062\u590d\u4f7f q \u53ef\u89e3\u51b3&#xff08;\u4f46\u4e0d\u5b9e\u73b0\u5b83&#xff09;\u6240\u9700\u7684\u4f9d\u8d56\u9879\u3002\u6211\u4eec\u7528\u00a0E^\u00a0\u8868\u793a\u6700\u7ec8\u8865\u5168\u7684\u5de5\u4f5c\u7a7a\u95f4\u3002\u6211\u4eec\u5c06\u6b64\u9636\u6bb5\u5e94\u7528\u4e8e\u6240\u6709\u56de\u653e\u7684\u5de5\u4f5c\u7a7a\u95f4&#xff0c;\u5176\u5bf9\u5de5\u4f5c\u7a7a\u95f4\u590d\u6742\u5ea6\u7684\u5f71\u54cd\u8be6\u89c1 \u00a74\u3002<\/p>\n<p>\u9636\u6bb5 3&#xff1a;\u73af\u5883\u8fc7\u6ee4\u3002\u00a0\u4e00\u4e2a\u8865\u5168\u7684\u5de5\u4f5c\u7a7a\u95f4\u53ea\u6709\u5728\u5176\u66b4\u9732\u8db3\u591f\u7684\u9879\u76ee\u4e0a\u4e0b\u6587\u4ee5\u652f\u6301\u5176\u4efb\u52a1\u65f6\u624d\u662f\u6709\u7528\u7684\u3002\u4e00\u4e2a\u667a\u80fd\u4f53\u8bc4\u5224\u5458\u4f7f\u7528\u53ea\u8bfb\u7684 shell \u548c\u6587\u4ef6\u5de5\u5177\u68c0\u67e5\u6bcf\u4e2a\u00a0E^&#xff0c;\u5e76\u6839\u636e\u6062\u590d\u7684\u4efb\u52a1 q&#xff0c;\u6807\u8bb0\u5de5\u4f5c\u7a7a\u95f4\u4e3a\u5145\u5206\u6216\u4e0d\u5145\u5206&#xff0c;\u4f9d\u636e\u662f\u5176\u6e90\u4ee3\u7801\u3001\u914d\u7f6e\u3001\u6570\u636e\u548c\u7ed3\u6784\u662f\u5426\u4e3a\u6709\u80fd\u529b\u7684\u667a\u80fd\u4f53\u63d0\u4f9b\u4e86\u8db3\u591f\u7684\u4e0a\u4e0b\u6587\u6765\u5904\u7406\u8be5\u4efb\u52a1 (\u00a7 B.3)\u3002\u53ea\u6709\u5145\u5206\u7684\u5de5\u4f5c\u7a7a\u95f4\u88ab\u4fdd\u7559\u7528\u4e8e\u4e0b\u6e38\u4efb\u52a1\u751f\u6210&#xff0c;\u6bcf\u4e2a\u9636\u6bb5\u7684\u5145\u5206\u7387\u548c\u6700\u7ec8\u8ba1\u6570\u5728 \u00a74 \u4e2d\u62a5\u544a\u3002<\/p>\n<p>\u6bcf\u4e2a\u91cd\u5efa\u7684\u5de5\u4f5c\u7a7a\u95f4\u8fd0\u884c\u5728\u6807\u51c6\u5316\u7684 ubuntu:24.04 \u5bb9\u5668\u4e2d&#xff0c;\u5177\u6709\u7f51\u7edc\u8bbf\u95ee\u6743\u9650\u3002\u4e0e\u7279\u5b9a\u4e8e\u4ee3\u7801\u4ed3\u5e93\u7684\u955c\u50cf\u76f8\u6bd4&#xff0c;\u8fd9\u79cd\u8bbe\u8ba1\u964d\u4f4e\u4e86\u6210\u672c\u5e76\u7b80\u5316\u4e86\u90e8\u7f72&#xff0c;\u5c3d\u7ba1\u5148\u524d\u5de5\u4f5c\u62a5\u544a\u79f0\u5176\u89e3\u51b3\u7387\u7565\u6709\u964d\u4f4e (Zeng et al., 2026)\u3002<\/p>\n<h4>3.2 \u91cd\u65b0\u67e5\u8be2<\/h4>\n<p>\u867d\u7136\u91cd\u5efa\u4ea7\u751f\u4e86\u53ef\u6267\u884c\u73af\u5883&#xff0c;\u4f46\u91cd\u65b0\u67e5\u8be2\u51b3\u5b9a\u4e86\u5982\u4f55\u6709\u6548\u5730\u5229\u7528\u5176\u6f5c\u5728\u80fd\u529b\u7a7a\u95f4\u3002\u6211\u4eec\u5f15\u5165\u4e86\u56db\u79cd\u4e92\u8865\u7684\u91cd\u65b0\u67e5\u8be2\u673a\u5236&#xff0c;\u5728\u6574\u7bc7\u8bba\u6587\u4e2d\u5206\u522b\u8868\u793a\u4e3a\u610f\u56fe\u6062\u590d\u3001\u5355\u5de5\u4f5c\u7a7a\u95f4\u3001\u8de8\u5de5\u4f5c\u7a7a\u95f4\u548c\u591a\u8f6e\u3002\u610f\u56fe\u6062\u590d\u91cd\u5efa\u6e90\u4efb\u52a1&#xff0c;\u800c\u5355\u5de5\u4f5c\u7a7a\u95f4\u5728\u5355\u4e2a\u5de5\u4f5c\u7a7a\u95f4\u5185\u5408\u6210\u65b0\u4efb\u52a1\u3002\u8de8\u5de5\u4f5c\u7a7a\u95f4\u901a\u8fc7\u8fde\u63a5\u76f8\u5173\u5de5\u4f5c\u7a7a\u95f4\u63d0\u4f9b\u5e7f\u5ea6&#xff0c;\u591a\u8f6e\u901a\u8fc7\u5c06\u521d\u59cb\u67e5\u8be2\u6269\u5c55\u4e3a\u4ea4\u4e92\u5f0f\u4f1a\u8bdd\u6765\u63d0\u4f9b\u6df1\u5ea6\u3002<\/p>\n<li>\n<p>\u73af\u5883\u91cd\u5efa<\/p>\n<\/li>\n<li>\n<p>\u91cd\u65b0\u67e5\u8be2\u65b9\u6cd5<\/p>\n<\/li>\n<li>\n<p>\u667a\u80fd\u4f53\u8f68\u8ff9<\/p>\n<ul>\n<li>\n<p>\u8bfb\u53d6 config\/config.json<\/p>\n<\/li>\n<li>\n<p>\u8bfb\u53d6 data\/input\/sample.txt<\/p>\n<\/li>\n<li>\n<p>\u5199\u5165 config\/settings.yaml<\/p>\n<\/li>\n<li>\n<p>\u5199\u5165 build\/linux_build.sh<\/p>\n<\/li>\n<li>\n<p>\u5199\u5165 process_data.py<\/p>\n<\/li>\n<li>\n<p>\u6267\u884c .\/build &amp;&amp; pytest<\/p>\n<\/li>\n<li>\n<p>\u5199\u5165 test_report.json<\/p>\n<\/li>\n<\/ul>\n<p>\u9636\u6bb5 1&#xff1a;\u786e\u5b9a\u6027\u56de\u653e \u56de\u653e\u7684\u5de5\u4f5c\u7a7a\u95f4\u00a0E\u02c9E\u02c9<\/p>\n<ul>\n<li>\n<p>\/app<\/p>\n<ul>\n<li>\n<p>build\/linux_build.sh<\/p>\n<\/li>\n<li>\n<p>config\/config.json<\/p>\n<\/li>\n<li>\n<p>data\/input\/sample1-3.txt<\/p>\n<\/li>\n<li>\n<p>process_data.py<\/p>\n<\/li>\n<li>\n<p>test_report.json<\/p>\n<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<p>\u9636\u6bb5 2&#xff1a;\u667a\u80fd\u4f53\u8865\u5168 \u8865\u5168\u7684\u5de5\u4f5c\u7a7a\u95f4\u00a0E\u02c9E\u02c9<\/p>\n<ul>\n<li>\n<p>\/app<\/p>\n<ul>\n<li>\n<p>\u90e8\u7f72\u914d\u7f6e &#043; \u6a21\u5f0f &#043; \u811a\u672c &#043; Makefile &#043; pyproject.toml &#043;<\/p>\n<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<p>\u9636\u6bb5 3&#xff1a;\u73af\u5883\u8fc7\u6ee4 \u4efb\u52a1\u5145\u5206\u7684\u5de5\u4f5c\u7a7a\u95f4<\/p>\n<\/li>\n<p>\u610f\u56fe\u6062\u590d\u3002\u00a0\u5c06\u4e00\u4e2a\u6216\u591a\u4e2a\u6e90\u7528\u6237\u8bf7\u6c42\u5408\u5e76\u4e3a\u4e00\u4e2a\u72ec\u7acb\u7684\u4efb\u52a1\u3002<\/p>\n<p>\u6211\u4eec\u5c06\u6bcf\u6761\u6e90\u8f68\u8ff9\u89c4\u8303\u5316\u4e3a\u7528\u6237\u8bf7\u6c42\u3001\u667a\u80fd\u4f53\u884c\u52a8\u548c\u6587\u4ef6\u66f4\u6539\u7684\u65f6\u95f4\u6d41\u3002\u5bf9\u4e8e\u5355\u8f6e\u8f68\u8ff9&#xff0c;\u552f\u4e00\u7684\u5b9e\u8d28\u6027\u7528\u6237\u8bf7\u6c42\u76f4\u63a5\u5b9a\u4e49\u4efb\u52a1\u3002\u5bf9\u4e8e\u591a\u8f6e\u8f68\u8ff9&#xff0c;\u7b2c\u4e00\u4e2a\u5b9e\u8d28\u6027\u7528\u6237\u8bf7\u6c42\u786e\u5b9a\u4efb\u52a1\u4e3b\u9898&#xff0c;\u800c\u540e\u7eed\u8bf7\u6c42\u5728\u5b83\u4eec\u6f84\u6e05\u3001\u7ea6\u675f\u6216\u6269\u5c55\u540c\u4e00\u4efb\u52a1\u65f6\u88ab\u7eb3\u5165&#xff1b;\u4e0d\u76f8\u5173\u7684\u4efb\u52a1\u8f6c\u79fb\u88ab\u6392\u9664\u3002\u6211\u4eec\u4f7f\u7528\u667a\u80fd\u4f53\u884c\u52a8\u548c\u6587\u4ef6\u8bc1\u636e\u6765\u89e3\u91ca\u8bf7\u6c42&#xff0c;\u4f46\u53ea\u4fdd\u7559\u7528\u6237\u9648\u8ff0\u7684\u9700\u6c42 (\u00a7 C.1)\u3002<\/p>\n<p>\u5355\u5de5\u4f5c\u7a7a\u95f4\u5408\u6210\u3002\u00a0\u4ece\u5355\u4e2a\u91cd\u5efa\u7684\u5de5\u4f5c\u7a7a\u95f4\u63a8\u5bfc\u65b0\u7684\u67e5\u8be2\u3002<\/p>\n<p>\u4e00\u4e2a\u79bb\u7ebf\u751f\u6210\u5668\u68c0\u67e5\u6bcf\u4e2a\u5de5\u4f5c\u7a7a\u95f4&#xff0c;\u5e76\u5728\u843d\u5730\u6027\u3001\u7ed3\u6784\u591a\u6837\u6027\u548c\u53ef\u9a8c\u8bc1\u6027\u7ea6\u675f\u4e0b\u5408\u6210\u4e94\u4e2a\u72ec\u7acb\u7684\u5019\u9009\u4efb\u52a1\u3002\u6211\u4eec\u4e3a\u6bcf\u4e2a\u73af\u5883\u968f\u673a\u9009\u62e9\u4e00\u4e2a\u6709\u6548\u5019\u9009\u7528\u4e8e\u5c55\u5f00\u548c\u9a8c\u8bc1\u3002<\/p>\n<p>\u5e7f\u5ea6\u6269\u5c55\u3002\u00a0\u5728\u5355\u4e2a\u67e5\u8be2\u4e2d\u8fde\u63a5\u76f8\u5173\u5de5\u4f5c\u7a7a\u95f4\u3002<\/p>\n<p>\u8de8\u5de5\u4f5c\u7a7a\u95f4\u5408\u6210\u3002\u00a0\u4e3a\u4e86\u5408\u6210\u8de8\u8d8a\u591a\u4e2a\u4ee3\u7801\u5e93\u7684\u4efb\u52a1&#xff0c;\u6211\u4eec\u53d1\u73b0\u6062\u590d\u73af\u5883\u4e4b\u95f4\u7684\u65b9\u5411\u6027\u4f9d\u8d56\u5173\u7cfb\u3002\u4e00\u4e2a\u667a\u80fd\u4f53\u9996\u5148\u5206\u6790\u6bcf\u4e2a\u5de5\u4f5c\u7a7a\u95f4\u7684\u6280\u672f\u9886\u57df\u548c\u5df2\u5b9e\u73b0\u80fd\u529b\u3002\u7136\u540e&#xff0c;\u6211\u4eec\u901a\u8fc7 TF-IDF \u6700\u8fd1\u90bb\u641c\u7d22\u68c0\u7d22\u5019\u9009\u5bf9&#xff0c;\u5e76\u91c7\u7528 LLM \u8bc4\u5224\u5458\u6765\u8bc6\u522b\u65b9\u5411\u6027\u4f9d\u8d56\u8fb9&#xff0c;\u5373\u76ee\u6807\u5de5\u4f5c\u7a7a\u95f4\u7f3a\u5c11\u53c2\u8003\u5de5\u4f5c\u7a7a\u95f4\u4e2d\u5df2\u5b9e\u73b0\u7684\u67d0\u9879\u80fd\u529b\u3002<\/p>\n<p>\u6211\u4eec\u4e3a\u6bcf\u4e00\u5bf9\u5206\u914d\u6070\u597d\u4e00\u4e2a\u4efb\u52a1\u3002\u6bcf\u4e2a\u8de8\u5de5\u4f5c\u7a7a\u95f4\u4efb\u52a1\u5c06\u4e00\u4e2a\u53ef\u5199\u7684\u76ee\u6807\u5de5\u4f5c\u7a7a\u95f4\u4e0e\u4e00\u4e2a\u4ee5\u53ea\u8bfb\u65b9\u5f0f\u6302\u8f7d\u5728\u5355\u72ec\u8def\u5f84\u7684\u53c2\u8003\u5de5\u4f5c\u7a7a\u95f4\u914d\u5bf9\u3002\u4efb\u52a1\u751f\u6210\u5668\u786e\u8ba4\u529f\u80fd\u5dee\u8ddd\u662f\u771f\u5b9e\u7684&#xff0c;\u5e76\u6307\u5b9a\u76ee\u6807\u4e2d\u5f25\u5408\u8be5\u5dee\u8ddd\u7684\u53ef\u89c2\u5bdf\u884c\u4e3a&#xff0c;\u8fd9\u4e9b\u884c\u4e3a\u53ef\u901a\u8fc7\u786e\u5b9a\u6027\u7684\u672c\u5730\u547d\u4ee4\u8fdb\u884c\u9a8c\u8bc1\u3002\u67e5\u8be2\u4ec5\u63d0\u4f9b\u8fd9\u4e9b\u76ee\u6807\u884c\u4e3a\u548c\u53c2\u8003\u7684\u6302\u8f7d\u8def\u5f84&#xff0c;\u800c\u4e0d\u63d0\u4f9b\u5176\u5185\u90e8\u7ec6\u8282&#xff0c;\u4f7f\u53c2\u8003\u6210\u4e3a\u771f\u6b63\u7684\u4f9d\u8d56\u9879\u3002\u56e0\u6b64&#xff0c;\u6c42\u89e3\u8005\u5fc5\u987b\u81ea\u884c\u5bfc\u822a\u3001\u5185\u5316\u5e76\u9002\u914d\u53c2\u8003\u5b9e\u73b0\u3002<\/p>\n<p>\u6df1\u5ea6\u6269\u5c55\u3002\u00a0\u901a\u8fc7\u7528\u6237\u667a\u80fd\u4f53\u7684\u540e\u7eed\u95ee\u9898\u6269\u5c55\u5df2\u89e3\u51b3\u7684\u7ec8\u7aef\u4efb\u52a1\u3002<\/p>\n<p>\u591a\u8f6e\u7528\u6237\u67e5\u8be2\u3002\u00a0\u4ece\u4e00\u4e2a\u521d\u59cb\u7684\u7ec8\u7aef\u67e5\u8be2\u5f00\u59cb&#xff0c;\u6211\u4eec\u5728\u7f16\u7801\u8f6e\u6b21\u4e4b\u95f4\u4fdd\u7559\u5de5\u4f5c\u7a7a\u95f4&#xff0c;\u5e76\u5728\u521d\u59cb\u54cd\u5e94\u4e4b\u540e\u5f15\u5165\u4e00\u4e2a\u7528\u6237\u667a\u80fd\u4f53\u3002\u5ef6\u7eed\u8fc7\u7a0b\u91c7\u7528\u4e24\u79cd\u534f\u8c03\u673a\u5236\u4ee5\u786e\u4fdd\u8fde\u8d2f\u4e14\u771f\u5b9e\u7684\u591a\u8f6e\u8f68\u8ff9&#xff1a;<\/p>\n<li>\n<p>\u6f14\u8fdb\u7684\u4efb\u52a1\u89c4\u683c\u8bf4\u660e\u3002\u00a0\u7528\u6237\u667a\u80fd\u4f53\u7ef4\u62a4\u4e00\u4e2a\u660e\u786e\u7684\u9700\u6c42\u8ddf\u8e2a\u5668&#xff0c;\u8bb0\u5f55\u6d3b\u8dc3\u7684\u3001\u5df2\u6ee1\u8db3\u7684\u548c\u5df2\u66f4\u65b0\u7684\u9700\u6c42\u3002\u5728\u6bcf\u6b21\u540e\u7eed\u884c\u52a8\u4e4b\u524d&#xff0c;\u5b83\u901a\u8fc7\u6dfb\u52a0\u3001\u4fee\u6539\u6216\u66ff\u6362\u7ea6\u675f\u6765\u66f4\u65b0\u6b64\u8ddf\u8e2a\u5668&#xff0c;\u968f\u7740\u5de5\u4f5c\u7a7a\u95f4\u7684\u6f14\u5316\u4fdd\u6301\u4e0a\u4e0b\u6587\u8fde\u8d2f\u6027\u3002<\/p>\n<\/li>\n<li>\n<p>\u8f6e\u6b21\u7ea7\u9a8c\u8bc1\u548c\u53cd\u9988\u3002\u00a0\u5728\u6bcf\u4e2a\u6269\u5c55\u8f6e\u6b21&#xff0c;\u7528\u6237\u667a\u80fd\u4f53\u63d0\u4ea4\u66f4\u65b0\u540e\u7684\u89c4\u683c\u8bf4\u660e&#xff0c;\u63d0\u793a\u4e00\u4e2a\u81ea\u52a8\u5316\u9a8c\u8bc1\u5668\u5728\u7f16\u7801\u667a\u80fd\u4f53\u884c\u52a8\u4e4b\u524d\u7f16\u5199\u8f6e\u6b21\u7ea7\u7684\u9a8c\u6536\u6d4b\u8bd5\u3002\u6536\u5230\u667a\u80fd\u4f53\u7684\u54cd\u5e94\u540e&#xff0c;\u6d4b\u8bd5\u8fd0\u884c\u5668\u8bc4\u4f30\u65b0\u6807\u51c6\u548c\u6d3b\u8dc3\u7684\u56de\u5f52\u68c0\u67e5\u3002\u6c42\u89e3\u5668\u667a\u80fd\u4f53\u4e25\u683c\u4e0e\u6d4b\u8bd5\u811a\u672c\u548c\u56de\u6eaf\u4fe1\u606f\u9694\u79bb&#xff1b;\u7528\u6237\u667a\u80fd\u4f53\u89e3\u91ca\u7ed3\u6784\u5316\u7684\u6d4b\u8bd5\u7ed3\u679c&#xff0c;\u5e76\u5c06\u4efb\u4f55\u5931\u8d25\u8f6c\u5316\u4e3a\u81ea\u7136\u7684\u3001\u7528\u6237\u53ef\u89c2\u5bdf\u7684\u62b1\u6028\u3002\u4e2d\u95f4\u5931\u8d25\u4fdd\u7559\u5728\u5bf9\u8bdd\u5386\u53f2\u4e2d&#xff0c;\u4e3a\u9519\u8bef\u8bca\u65ad\u548c\u6062\u590d\u63d0\u4f9b\u73b0\u5b9e\u7684\u76d1\u7763\u3002<\/p>\n<\/li>\n<p>\u5728\u5404\u8f6e\u6b21\u4e2d&#xff0c;\u7528\u6237\u667a\u80fd\u4f53\u7684\u8bf7\u6c42\u5206\u4e3a\u4e09\u79cd\u4ea4\u4e92\u98ce\u683c&#xff1a;(i) \u7279\u6027\u6269\u5c55&#xff0c;\u5728\u6210\u529f\u9a8c\u8bc1\u540e\u5f15\u5165\u65b0\u7684\u843d\u5730\u9700\u6c42&#xff1b;(ii) \u7279\u6027\u4fee\u8ba2&#xff0c;\u7531\u6d4b\u8bd5\u5931\u8d25\u89e6\u53d1&#xff0c;\u57fa\u4e8e\u89c2\u5bdf\u5230\u7684\u884c\u4e3a\u8981\u6c42\u9519\u8bef\u4fee\u590d&#xff1b;\u4ee5\u53ca (iii) \u7279\u6027\u51b2\u7a81&#xff0c;\u4fee\u6539\u6216\u8986\u76d6\u5148\u524d\u7684\u89c4\u683c\u8bf4\u660e\u4ee5\u53cd\u6620\u53d8\u5316\u7684\u7528\u6237\u610f\u56fe\u3002\u6bcf\u79cd\u8bf7\u6c42\u7684\u98ce\u683c\u81ea\u7136\u5730\u9075\u5faa\u8f6e\u6b21\u7ed3\u679c\u548c\u4f1a\u8bdd\u4e0a\u4e0b\u6587&#xff0c;\u89c2\u5bdf\u5230\u7684\u5206\u5e03\u60c5\u51b5\u5728 \u00a7 E.3&#xff08;\u8868 17&#xff09;\u4e2d\u62a5\u544a\u3002\u4f1a\u8bdd\u6301\u7eed\u6700\u591a\u516d\u4e2a\u540e\u7eed\u8f6e\u6b21&#xff0c;\u6216\u76f4\u5230\u53d1\u51fa\u7ec8\u6b62\u6807\u8bb0\u3002<\/p>\n<p>\u56fe 3 \u62a5\u544a\u4e86\u7ecf\u8fc7 \u00a73.3 \u4e2d\u8f6e\u6b21\u7ea7\u9009\u62e9\u540e\u7684\u5ef6\u7eed\u6570\u636e\u7684\u901a\u8fc7\/\u5931\u8d25\u6a21\u5f0f&#xff1a;\u4fdd\u7559\u7684 3,079 \u6761\u8bb0\u5f55\u5e73\u5747\u6709 4.51 \u8f6e&#xff0c;\u5176\u4e2d 69.6% \u5305\u542b\u4e00\u4e2a\u4f1a\u8bdd\u968f\u540e\u4fee\u590d\u7684\u5931\u8d25&#xff0c;\u4fdd\u7559\u4e86\u6062\u590d\u76d1\u7763\u4fe1\u606f\u3002<\/p>\n<p class=\"img-center\"><img loading=\"lazy\" decoding=\"async\" alt=\"\" height=\"898\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260905232627-6a9ca52309213.png\" width=\"890\" \/><\/p>\n<p>\u56fe 3&#xff1a;\u591a\u8f6e\u4f1a\u8bdd\u4e2d\u7684\u901a\u8fc7\/\u5931\u8d25\u6a21\u5f0f\u3002<\/p>\n<h4>3.3 \u9a8c\u8bc1\u4e0e\u8fc7\u6ee4<\/h4>\n<p>\u667a\u80fd\u4f53\u9a8c\u8bc1\u5668\u6784\u5efa\u3002\u00a0\u6bcf\u4e2a\u5355\u5de5\u4f5c\u7a7a\u95f4\u6216\u8de8\u5de5\u4f5c\u7a7a\u95f4\u4efb\u52a1\u90fd\u914d\u6709\u4e00\u4e2a\u7531\u4e13\u7528\u667a\u80fd\u4f53\u5728\u76ee\u6807\u5bb9\u5668\u5185\u7f16\u5199\u7684\u53ef\u6267\u884c\u9a8c\u8bc1\u5668\u3002\u6839\u636e\u4efb\u52a1\u89c4\u683c\u8bf4\u660e\u548c\u5de5\u4f5c\u7a7a\u95f4\u6587\u4ef6&#xff0c;\u9a8c\u8bc1\u5668\u667a\u80fd\u4f53\u901a\u8fc7\u8fed\u4ee3\u7684\u672c\u5730\u6267\u884c\u5236\u4f5c\u4e00\u4e2a\u81ea\u5305\u542b\u7684 pytest \u5957\u4ef6&#xff0c;\u4e25\u683c\u8bc4\u4f30\u63d0\u793a\u4e2d\u6307\u5b9a\u7684\u516c\u5171\u63a5\u53e3\u548c\u9884\u671f\u884c\u4e3a (\u00a7 D)\u3002<\/p>\n<p>\u89e3\u51b3\u65b9\u6848\u5c55\u5f00\u3002\u00a0\u5019\u9009\u89e3\u51b3\u65b9\u6848\u4f7f\u7528\u6559\u5e08\u6a21\u578b\u5728\u4efb\u52a1\u5bb9\u5668\u5185\u7684 Claude Code \u6846\u67b6 (Anthropic, 2026) \u4e2d\u5c55\u5f00\u3002\u5c55\u5f00\u4f7f\u7528\u6e29\u5ea6 1.0&#xff0c;top-p &#061; 0.95&#xff0c;\u4ea4\u9519\u601d\u8003&#xff0c;\u4e0a\u4e0b\u6587\u7a97\u53e3\u4e3a 256k token&#xff0c;\u5355\u6b21\u54cd\u5e94\u9650\u5236\u4e3a 65,536 token&#xff0c;\u5728 176k token \u65f6\u89e6\u53d1\u4e3b\u52a8\u603b\u7ed3&#xff0c;\u6700\u591a 500 \u4e2a\u667a\u80fd\u4f53\u8f6e\u6b21&#xff0c;\u58c1\u949f\u8d85\u65f6\u4e3a\u56db\u5c0f\u65f6\u3002\u4efb\u52a1\u73af\u5883\u63d0\u4f9b\u5bb9\u5668\u5316\u7684\u7f51\u7edc\u8bbf\u95ee\u4ee5\u83b7\u53d6\u7f3a\u5931\u7684\u4f9d\u8d56\u9879 (\u00a73.1)\u3002<\/p>\n<p>\u9a8c\u8bc1\u548c\u6570\u636e\u9009\u62e9\u3002\u00a0\u5bf9\u4e8e\u5355\u5de5\u4f5c\u7a7a\u95f4\u548c\u8de8\u5de5\u4f5c\u7a7a\u95f4&#xff0c;\u7f16\u5199\u7684\u6d4b\u8bd5\u5957\u4ef6\u5728\u5c55\u5f00\u540e\u7684\u6700\u7ec8\u5de5\u4f5c\u7a7a\u95f4\u72b6\u6001\u4e0b\u8fd0\u884c&#xff0c;\u53ea\u6709\u5f53\u6240\u6709\u6d4b\u8bd5\u90fd\u901a\u8fc7\u65f6&#xff0c;\u8f68\u8ff9\u624d\u88ab\u63a5\u53d7\u3002\u591a\u8f6e\u6570\u636e\u5728\u8f6e\u6b21\u7ea7\u522b\u8fdb\u884c\u9009\u62e9&#xff1a;\u6211\u4eec\u4ece\u7ec8\u6b62\u7684\u4f1a\u8bdd\u4e2d\u4fee\u526a\u8fde\u7eed\u5931\u8d25\u7684\u8f6e\u6b21\u540e\u7f00&#xff0c;\u5e76\u4e14\u4ec5\u5f53\u4f1a\u8bdd\u5305\u542b\u81f3\u5c11\u4e24\u4e2a\u901a\u8fc7\u9a8c\u8bc1\u7684\u8f6e\u6b21\u65f6\u624d\u4fdd\u7559\u5b83\u3002\u5728\u6210\u529f\u6062\u590d\u4e4b\u524d\u7684\u4e2d\u95f4\u5931\u8d25\u88ab\u4fdd\u7559&#xff0c;\u4e3a\u9519\u8bef\u8bca\u65ad\u548c\u6062\u590d\u63d0\u4f9b\u76d1\u7763\u3002\u63a5\u53d7\u7684\u8f68\u8ff9\u88ab\u683c\u5f0f\u5316\u4e3a\u591a\u8f6e SFT \u6f14\u793a&#xff0c;\u5e76\u9488\u5bf9\u8bc4\u4f30\u57fa\u51c6\u8fdb\u884c\u4e25\u683c\u53bb\u6c61 (\u00a75)\u3002<\/p>\n<p>\u4e3a\u4e86\u4e3a\u4e0a\u8ff0\u673a\u5236\u63d0\u4f9b\u5177\u4f53\u76f4\u89c2\u7684\u7406\u89e3&#xff0c;\u6211\u4eec\u5728\u9644\u5f55\u4e2d\u8be6\u7ec6\u4ecb\u7ecd\u4e86\u51e0\u4e2a\u7efc\u5408\u6848\u4f8b\u7814\u7a76\u3002\u00a7 E.1 \u8ffd\u8e2a\u4e86\u4e00\u4e2a\u8f68\u8ff9\u5230\u67e5\u8be2\u7684\u793a\u4f8b&#xff0c;\u00a7 E.2 \u8ddf\u8e2a\u4e86\u4e00\u4e2a\u8de8\u5de5\u4f5c\u7a7a\u95f4\u4ee3\u7801\u4ed3\u5e93\u96c6\u4ece\u5019\u9009\u6765\u6e90\u5230\u9009\u5b9a\u4efb\u52a1\u7684\u8fc7\u7a0b&#xff0c;\u00a7 E.3 \u8ddf\u8e2a\u4e86\u4e00\u4e2a\u591a\u8f6e\u5ef6\u7eed\u4ece\u9a8c\u8bc1\u5668\u9a71\u52a8\u7684\u4fee\u8ba2\u5230\u53d7\u63a7\u9700\u6c42\u53d8\u66f4\u7684\u8fc7\u7a0b&#xff08;\u56fe 8&#xff09;\u3002\u00a74 \u62a5\u544a\u6570\u636e\u96c6\u5927\u5c0f&#xff0c;\u00a7F \u62a5\u544a\u4ea4\u4e92\u7edf\u8ba1\u4fe1\u606f\u3002<\/p>\n<h3>4 \u89c4\u6a21\u5316\u6570\u636e\u6784\u5efa<\/h3>\n<h4>4.1 \u79cd\u5b50\u9009\u62e9<\/h4>\n<p>\u6211\u4eec\u4ece\u4e0d\u540c\u7684\u7ec8\u7aef\u98ce\u683c CLI \u548c\u8f6f\u4ef6\u5de5\u7a0b\u8bed\u6599\u5e93\u4e2d\u83b7\u53d6\u539f\u59cb\u667a\u80fd\u4f53\u8f68\u8ff9\u3002\u5982\u679c\u6267\u884c\u8f68\u8ff9\u7ed3\u675f\u65f6\u89c2\u5bdf\u5230\u7684\u5de5\u4f5c\u7a7a\u95f4\u72b6\u6001\u5305\u542b\u81f3\u5c11 5 \u4e2a\u6587\u4ef6\u548c 100 \u884c\u4ee3\u7801&#xff0c;\u5219\u8be5\u8f68\u8ff9\u88ab\u4fdd\u7559\u4e3a\u53ef\u884c\u79cd\u5b50\u3002\u4e3a\u9632\u6b62\u6570\u636e\u6cc4\u6f0f&#xff0c;\u6211\u4eec\u4e25\u683c\u8fc7\u6ee4\u6389\u6240\u6709\u6e90\u81ea Terminal-Bench \u7684\u6570\u636e\u96c6\u3002\u5b8c\u6574\u7684\u6765\u6e90\u7ec6\u5206\u548c\u9009\u62e9\u7edf\u8ba1\u4fe1\u606f\u8be6\u89c1 \u00a7 A\u3002<\/p>\n<h4>4.2 \u91cd\u5efa\u7edf\u8ba1<\/h4>\n<p>\u4ece\u6e90\u8f68\u8ff9\u6c60\u4e2d&#xff0c;\u91cd\u5efa\u6d41\u6c34\u7ebf\u4ea7\u751f\u4e86 68,263 \u4e2a\u91cd\u5efa\u73af\u5883\u3002\u5982\u56fe 4 \u6240\u793a&#xff0c;\u56de\u653e\u7684\u7ec8\u7aef\u5de5\u4f5c\u7a7a\u95f4\u6700\u521d\u5305\u542b\u5f88\u5c11\u6587\u4ef6&#xff0c;\u56e0\u4e3a\u5728\u539f\u59cb\u5c55\u5f00\u671f\u95f4\u751f\u6210\u7684\u89e3\u51b3\u65b9\u6848\u6587\u4ef6\u88ab\u4fdd\u7559\u3002\u667a\u80fd\u4f53\u8865\u5168\u663e\u8457\u4e30\u5bcc\u4e86\u8fd9\u4e9b\u73af\u5883&#xff0c;\u5c06\u5e73\u5747\u6587\u4ef6\u6570\u4ece 2.9 \u589e\u52a0\u5230 22.4\u3002\u867d\u7136 SWE \u79cd\u5b50\u6e90\u81ea\u66f4\u4e30\u5bcc\u7684\u4ee3\u7801\u4ed3\u5e93\u72b6\u6001&#xff0c;\u4f46\u8865\u5168\u540c\u6837\u6269\u5c55\u4e86\u5176\u4e0a\u4e0b\u6587\u5e7f\u5ea6&#xff08;\u5b8c\u6574\u7684\u590d\u6742\u5ea6\u5206\u5e03\u89c1\u8868 13&#xff09;\u3002<\/p>\n<p class=\"img-center\"><img loading=\"lazy\" decoding=\"async\" alt=\"\" height=\"570\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260905232627-6a9ca52341ad8.png\" width=\"1734\" \/><\/p>\n<p>\u56fe 4&#xff1a;\u667a\u80fd\u4f53\u8865\u5168\u524d\u540e\u7684\u5e73\u5747\u5de5\u4f5c\u7a7a\u95f4\u5927\u5c0f\u3002<\/p>\n<p>\u5728\u6c61\u67d3\u8fc7\u6ee4\u548c\u4ee3\u7801\u4ed3\u5e93\u7ea7\u53bb\u91cd\u4e4b\u540e&#xff0c;\u6211\u4eec\u5c06\u7b2c 3.1 \u8282\u4e2d\u7684\u9636\u6bb5 3 \u5145\u5206\u6027\u8bc4\u5224\u5668\u5e94\u7528\u4e8e\u5b58\u6d3b\u7684\u8865\u5168\u73af\u5883\u3002<\/p>\n<p>\u8868 2&#xff1a;\u5728\u6062\u590d\u610f\u56fe\u4e0b\u7684\u5de5\u4f5c\u7a7a\u95f4\u5145\u5206\u6027\u3002<\/p>\n<table>\n<tr>\u6c60# \u73af\u5883\u5145\u5206\u7387 (%)<\/tr>\n<tbody>\n<tr>\n<td><\/td>\n<td><\/td>\n<td>\u56de\u653e\u540e<\/td>\n<td>\u8865\u5168\u540e<\/td>\n<\/tr>\n<tr>\n<td>\u7ec8\u7aef<\/td>\n<td>38,294<\/td>\n<td>40.2<\/td>\n<td>93.5<\/td>\n<\/tr>\n<tr>\n<td>SWE<\/td>\n<td>1,900<\/td>\n<td>20.1<\/td>\n<td>77.1<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u5982\u8868 2 \u6240\u793a&#xff0c;\u4ec5\u786e\u5b9a\u6027\u56de\u653e\u5bf9\u7ec8\u7aef\u5de5\u4f5c\u7a7a\u95f4\u7684\u5145\u5206\u7387\u4e3a 40.2%&#xff0c;\u5bf9 SWE \u5de5\u4f5c\u7a7a\u95f4\u4e3a 20.1%\u3002\u667a\u80fd\u4f53\u8865\u5168\u663e\u8457\u63d0\u9ad8\u4e86\u8986\u76d6\u7387&#xff0c;\u5728\u4ee3\u7801\u4ed3\u5e93\u7ea7\u53bb\u91cd\u540e&#xff0c;\u5206\u522b\u5bf9 38,294 \u4e2a\u8bc4\u4f30\u7684\u7ec8\u7aef\u73af\u5883\u548c 1,900 \u4e2a SWE \u4ee3\u7801\u4ed3\u5e93&#xff08;\u6bcf\u4e2a\u4ee3\u7801\u4ed3\u5e93\u4e00\u4e2a\u4ee3\u8868\u6027\u91cd\u5efa&#xff09;\u8fbe\u5230\u4e86 93.5% \u548c 77.1%\u3002\u603b\u5171&#xff0c;\u4efb\u52a1\u5bfc\u5411\u7684\u8bc4\u4f30\u8bc6\u522b\u51fa 37,273 \u4e2a\u5b8c\u5168\u5145\u5206\u7684\u73af\u5883&#xff0c;\u9002\u7528\u4e8e\u4e0b\u6e38\u4efb\u52a1\u751f\u6210 (\u00a7 A)\u3002<\/p>\n<p class=\"img-center\"><img loading=\"lazy\" decoding=\"async\" alt=\"\" height=\"690\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260905232627-6a9ca523bf605.png\" width=\"1732\" \/><\/p>\n<p>\u56fe 5 \u5c55\u793a\u4e86\u91cd\u5efa\u7ec8\u7aef\u6c60\u7684\u591a\u6837\u6027&#xff1a;Python \u662f\u4e3b\u8981\u7f16\u7a0b\u8bed\u8a00 (84.7%)&#xff0c;\u5176\u6b21\u662f C&#043;&#043; \u548c C&#xff0c;\u6570\u636e\u5904\u7406\u3001DevOps \u548c\u5b89\u5168\u5de5\u4f5c\u8d1f\u8f7d\u5408\u8ba1\u5360\u6280\u672f\u9886\u57df\u7684 80% \u4ee5\u4e0a\u3002<\/p>\n<h4 style=\"background-color:transparent\">4.3 SFT \u8bed\u6599\u5e93\u7ec4\u6210<\/h4>\n<p>\u8fd9\u4e9b\u4efb\u52a1\u5145\u5206\u7684\u73af\u5883\u662f\u6211\u4eec\u56db\u79cd\u91cd\u65b0\u67e5\u8be2\u53d8\u4f53\u7684\u57fa\u7840&#xff1a;\u610f\u56fe\u6062\u590d\u91cd\u5efa\u539f\u59cb\u4efb\u52a1&#xff0c;\u5355\u5de5\u4f5c\u7a7a\u95f4\u4ea7\u751f\u65b0\u7684\u4ed3\u5e93\u5185\u4efb\u52a1&#xff0c;\u8de8\u5de5\u4f5c\u7a7a\u95f4\u53d1\u73b0\u4ed3\u5e93\u95f4\u7684\u4f9d\u8d56\u6311\u6218&#xff0c;\u591a\u8f6e\u5c06\u521d\u59cb\u67e5\u8be2\u6269\u5c55\u4e3a\u8fed\u4ee3\u4f1a\u8bdd\u3002\u5728\u6574\u4e2a\u7ec8\u7aef\u6c60\u4e2d&#xff0c;\u7ecf\u8fc7\u9a8c\u8bc1\u5668\u8fc7\u6ee4\u7684\u5355\u5de5\u4f5c\u7a7a\u95f4\u3001\u8de8\u5de5\u4f5c\u7a7a\u95f4\u548c\u591a\u8f6e\u5408\u6210\u4ea7\u751f\u4e86 31,977 \u4e2a SFT \u6f14\u793a&#xff08;25,386 \u4e2a\u5355\u5de5\u4f5c\u7a7a\u95f4&#xff0c;3,512 \u4e2a\u8de8\u5de5\u4f5c\u7a7a\u95f4&#xff0c;\u548c 3,079 \u4e2a\u591a\u8f6e\u8f68\u8ff9&#xff09;&#xff0c;\u603b\u8ba1\u7ea6 14.2 \u4ebf\u8bad\u7ec3 token\u3002\u4e2d\u4f4d\u6570\u4ea4\u4e92\u7edf\u8ba1\u4fe1\u606f\u5728 \u00a7F \u4e2d\u63d0\u4f9b\u3002<\/p>\n<h3>5 \u5b9e\u9a8c<\/h3>\n<h4>5.1 \u8bbe\u7f6e<\/h4>\n<p>\u8bad\u7ec3\u3002\u00a0\u6211\u4eec\u5728 Terminal-Universe \u6570\u636e\u4e0a\u5bf9 Qwen3.5-27B \u8fdb\u884c\u4e24\u8f6e SFT \u5fae\u8c03\u3002\u6211\u4eec\u4f7f\u7528\u6052\u5b9a\u7684 7\u00d710\u207b\u2076 \u5b66\u4e60\u7387&#xff0c;\u5168\u5c40\u6279\u6b21\u5927\u5c0f\u4e3a 256&#xff0c;\u5e8f\u5217\u957f\u5ea6\u4e3a 256k token\u3002\u8bad\u7ec3\u524d&#xff0c;\u6211\u4eec\u9488\u5bf9 Terminal-Bench \u4efb\u52a1\u8fd0\u884c 13-gram \u6c61\u67d3\u68c0\u67e5&#xff0c;\u5e76\u4ece\u6240\u6709\u56db\u79cd\u91cd\u65b0\u67e5\u8be2\u53d8\u4f53\u4e2d\u6392\u9664\u6e90\u81ea Terminal-Bench \u7684\u6570\u636e\u96c6\u3002<\/p>\n<p>\u5355\u8f6e\u8bc4\u4f30&#xff08;Terminal-Bench 2.0 \u548c 2.1&#xff09;\u3002\u00a0\u6211\u4eec\u4f7f\u7528 Claude Code&#xff08;\u7248\u672c 2.1.126&#xff09;\u548c Terminus2 (Merrill et al., 2026) \u53ca XML \u89e3\u6790\u5668\u8bc4\u4f30 Terminal-Bench\u3002\u8bc4\u4f30\u955c\u50cf\u4e86\u89e3\u51b3\u65b9\u6848\u5c55\u5f00 (\u00a73.3) \u7684\u89e3\u7801\u548c\u6267\u884c\u914d\u7f6e&#xff0c;\u91c7\u7528\u6e29\u5ea6 1.0&#xff0c;top-p &#061; 0.95&#xff0c;\u4ea4\u9519\u601d\u8003&#xff0c;256k \u4e0a\u4e0b\u6587\u7a97\u53e3&#xff0c;65,536 token \u8f6e\u6b21\u9650\u5236&#xff0c;176k token \u4e3b\u52a8\u603b\u7ed3&#xff0c;\u6700\u591a 500 \u4e2a\u667a\u80fd\u4f53\u8f6e\u6b21&#xff0c;\u4ee5\u53ca\u56db\u5c0f\u65f6\u58c1\u949f\u8d85\u65f6\u3002\u5bf9\u4e8e Claude Code&#xff0c;\u4ea4\u4e92\u5f0f\u548c\u7f51\u7edc\u68c0\u7d22\u5de5\u5177&#xff08;WebFetch&#xff0c; WebSearch, AskUserQuestion, EnterPlanMode, ExitPlanMode&#xff09;\u88ab\u7981\u7528\u3002\u6bcf\u4e2a\u5bb9\u5668\u8fd0\u884c\u5728 12 \u4e2a CPU \u6838\u5fc3\u548c 32 GiB \u5185\u5b58\u4e0a\u3002\u62a5\u544a\u7684\u5206\u6570\u4ee3\u8868\u516d\u6b21\u72ec\u7acb\u8fd0\u884c\u7684\u5e73\u5747\u901a\u8fc7\u7387\u3002Terminal-Bench 2.0 \u548c 2.1 \u4e4b\u95f4\u7684\u5dee\u5f02\u8bb0\u5f55\u5728\u57fa\u51c6\u66f4\u65b0\u4e2d\u3002<\/p>\n<p>\u591a\u8f6e\u8bc4\u4f30&#xff08;EvoCode-Bench v2&#xff09;\u3002\u00a0\u6211\u4eec\u5728 EvoCode-Bench v2 (Shen et al., 2026a) \u4e0a\u8bc4\u4f30\u6df1\u5ea6\u6269\u5c55&#xff0c;\u5b83\u5305\u542b 26 \u4e2a\u7f16\u7801\u4efb\u52a1\u548c 227 \u8f6e&#xff08;\u6bcf\u4e2a\u4efb\u52a1 5-15 \u4e2a\u8bf7\u6c42&#xff09;\u3002\u5de5\u4f5c\u7a7a\u95f4\u548c\u4f1a\u8bdd\u662f\u6301\u4e45\u7684&#xff0c;\u7d2f\u79ef\u9a8c\u8bc1\u5668\u68c0\u67e5\u5f53\u524d\u548c\u5148\u524d\u7684\u6d3b\u8dc3\u9700\u6c42\u3002\u6211\u4eec\u7684 Qwen3.5-27B \u68c0\u67e5\u70b9\u4f7f\u7528 Terminus2-XML&#xff0c;\u6e29\u5ea6 1.0&#xff0c;\u4ea4\u9519\u601d\u8003&#xff0c;256k token \u4e0a\u4e0b\u6587&#xff0c;\u6bcf\u8f6e 65,536 token \u9650\u5236&#xff0c;176k token \u4e3b\u52a8\u603b\u7ed3&#xff0c;\u6bcf\u4e2a\u8bf7\u6c42\u8f6e\u6b21\u6700\u591a 500 \u4e2a\u667a\u80fd\u4f53\u8f6e\u6b21&#xff0c;\u6bcf\u4e2a\u72b6\u6001\u5316\u4efb\u52a1\u603b\u58c1\u949f\u9650\u5236\u4e3a 10 \u5c0f\u65f6\u3002\u6bcf\u4e2a\u4efb\u52a1\u8bc4\u4f30\u56db\u6b21\u72ec\u7acb\u8fd0\u884c\u3002MT&#064;4 \u4f7f\u7528\u5931\u8d25\u505c\u6b62\u8bc4\u5206&#xff1a;\u4e00\u6b21\u8fd0\u884c\u4ec5\u5728\u7b2c\u4e00\u6b21\u5931\u8d25\u524d\u8fde\u7eed\u901a\u8fc7\u7684\u8f6e\u6b21\u83b7\u5f97\u5206\u6570&#xff0c;\u5e76\u4e14\u5982\u679c\u4efb\u4f55\u8fd0\u884c\u6210\u529f\u5230\u8fbe\u8be5\u4efb\u52a1-\u8f6e\u6b21&#xff0c;\u5219\u8be5\u4efb\u52a1-\u8f6e\u6b21\u88ab\u8ba1\u5206\u3002\u8be5\u5206\u6570\u5bf9\u6bcf\u4e2a\u4efb\u52a1\u5185\u5df2\u8ba1\u5206\u7684\u8f6e\u6b21\u8fdb\u884c\u5e73\u5747&#xff0c;\u7136\u540e\u8de8\u4efb\u52a1\u5e73\u5747\u3002\u6848\u4f8b\u5206\u6570\u5bf9\u4efb\u52a1\u5185\u7684\u9a8c\u8bc1\u5668\u6848\u4f8b\u901a\u8fc7\u7387\u8fdb\u884c\u5e73\u5747&#xff0c;\u7136\u540e\u8de8\u4efb\u52a1\u548c\u8fd0\u884c\u5e73\u5747&#xff0c;\u4ee5\u8861\u91cf\u90e8\u5206\u8fdb\u5c55\u3002<\/p>\n<h4>5.2 \u4e3b\u8981\u7ed3\u679c<\/h4>\n<p>\u8868 3 \u62a5\u544a\u4e86\u5355\u8f6e\u548c\u591a\u8f6e\u6027\u80fd\u3002\u5728\u5b8c\u6574\u6df7\u5408\u6570\u636e\u96c6&#xff08;\u7ecf\u8fc7\u9a8c\u8bc1\u5668\u8fc7\u6ee4\u7684\u5355\u5de5\u4f5c\u7a7a\u95f4\u3001\u8de8\u5de5\u4f5c\u7a7a\u95f4\u548c\u591a\u8f6e\u8f68\u8ff9&#xff0c;\u5171 32.0k \u6761\u8bb0\u5f55&#xff09;\u4e0a\u5fae\u8c03 Qwen3.5-27B&#xff0c;\u5728 Terminal-Bench 2.0 \u4e0a\u8fbe\u5230 52.8%&#xff0c;\u5728 Terminal-Bench 2.1 \u4e0a\u8fbe\u5230 58.1%&#xff08;\u5206\u522b\u6bd4\u57fa\u7840\u6a21\u578b\u63d0\u9ad8 11.2 \u548c 11.9 \u4e2a\u767e\u5206\u70b9&#xff09;\u5728 Terminus2-XML \u4e0b\u3002\u5728 Claude Code \u4e0b&#xff0c;\u5b83\u5728 Terminal-Bench 2.1 \u4e0a\u8fbe\u5230 58.2%&#xff08;\u6bd4\u57fa\u7840\u6a21\u578b\u63d0\u9ad8 10.4 \u4e2a\u767e\u5206\u70b9&#xff09;\u3002<\/p>\n<p>\u8868 3&#xff1a;Terminal-Bench \u548c EvoCode-Bench v2 \u7ed3\u679c (%)\u3002Terminal-Bench \u5217\u62a5\u544a\u5e73\u5747 Pass&#064;1\u3002\u7eff\u8272\u7c97\u4f53\u6807\u8bb0\u4efb\u52a1\u5408\u6210\u65b9\u6cd5\u4e2d\u7684\u6700\u4f73\u5206\u6570&#xff0c;\u84dd\u8272\u4e0b\u5212\u7ebf\u6807\u8bb0\u7b2c\u4e8c\u4f73\u5206\u6570\u3002\u5e26 \\\\* \u7684\u5206\u6570\u662f\u6211\u4eec\u6309\u7167 \u00a7G \u4e2d\u7684\u914d\u7f6e\u5728\u53d1\u5e03\u6a21\u578b\u4e0a\u6d4b\u91cf\u7684&#xff0c;\u5176\u4ed6\u5206\u6570\u53d6\u81ea\u76f8\u5e94\u62a5\u544a\u3002<\/p>\n<p class=\"img-center\"><img loading=\"lazy\" decoding=\"async\" alt=\"\" height=\"1068\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260905232628-6a9ca5245545f.png\" width=\"1710\" \/><\/p>\n<p>\u5728 EvoCode-Bench v2 \u4e0a&#xff0c;\u5b8c\u6574\u6df7\u5408\u6570\u636e\u96c6\u5c06 MT&#064;4 \u4ece 6.3 \u63d0\u9ad8\u5230 20.1&#xff0c;\u6848\u4f8b\u5206\u6570\u4ece 67.8 \u63d0\u9ad8\u5230 76.1&#xff0c;\u8868\u660e\u5b8c\u6574\u6df7\u5408\u6570\u636e\u96c6\u8bad\u7ec3\u4e5f\u63d0\u9ad8\u4e86\u5728\u5177\u6709\u7d2f\u79ef\u9700\u6c42\u7684\u6301\u4e45\u4efb\u52a1\u4e0a\u7684\u6027\u80fd\u3002<\/p>\n<h3>6 \u5206\u6790\u4e0e\u6d88\u878d\u7814\u7a76<\/h3>\n<p>\u6211\u4eec\u7684\u65b9\u6cd5\u5305\u542b\u591a\u4e2a\u8bbe\u8ba1\u9009\u62e9&#xff0c;\u672c\u8282\u5c06\u9694\u79bb\u6bcf\u4e2a\u9009\u62e9\u7684\u8d21\u732e\u3002\u6211\u4eec\u63d0\u51fa\u4ee5\u4e0b\u95ee\u9898&#xff1a;(1) \u91cd\u5efa\u73af\u5883\u5e76\u91cd\u65b0\u89e3\u51b3\u5b83\u662f\u5426\u771f\u7684\u6bd4\u5728\u539f\u59cb\u8f68\u8ff9\u4e0a\u8bad\u7ec3\u66f4\u597d (\u00a76.1)&#xff1f;(2) \u667a\u80fd\u4f53\u8865\u5168\u662f\u5426\u91cd\u8981&#xff0c;\u8fd8\u662f\u786e\u5b9a\u6027\u56de\u653e\u5c31\u8db3\u591f\u4e86 (\u00a76.2)&#xff1f;(3) \u9a8c\u8bc1\u5668\u8fc7\u6ee4\u662f\u5426\u503c\u5f97\u4e22\u5f03\u7684\u6570\u636e (\u00a76.3)&#xff1f;(4) \u5728\u56fa\u5b9a\u7684\u9884\u7b97\u4e0b&#xff0c;\u5e94\u5982\u4f55\u5728\u91cd\u65b0\u67e5\u8be2\u8f74\u2014\u2014\u5e7f\u5ea6 (\u00a76.4)\u3001\u6df1\u5ea6 (\u00a76.5) \u4ee5\u53ca\u66f4\u591a\u73af\u5883\u4e0e\u66f4\u591a\u67e5\u8be2\u4e4b\u95f4\u5206\u914d (\u00a76.6)&#xff1f;\u6211\u4eec\u6700\u540e\u6d4b\u8bd5\u4e86\u8be5\u6d41\u6c34\u7ebf\u662f\u5426\u80fd\u8fc1\u79fb\u5230\u7ec8\u7aef\u5de5\u4f5c\u7a7a\u95f4\u4e4b\u5916 (\u00a76.7)\u3002<\/p>\n<h4>6.1 \u4efb\u52a1\u91cd\u65b0\u89e3\u51b3<\/h4>\n<p>\u8fd9\u9879\u6d88\u878d\u7814\u7a76\u8be2\u95ee\u662f\u5426\u771f\u7684\u9700\u8981\u91cd\u5efa&#xff1a;\u91cd\u65b0\u89e3\u51b3\u6062\u590d\u7684\u4efb\u52a1\u662f\u5426\u6bd4\u7b80\u5355\u6a21\u4eff\u539f\u59cb\u8f68\u8ff9\u66f4\u597d&#xff1f;\u8868 4 \u5728 Terminal-Bench 2.1 \u4e0a\u6bd4\u8f83\u4e86\u8fd9\u4e24\u8005\u3002\u4e3a\u4e86\u4fdd\u6301\u6bd4\u8f83\u7684\u6e05\u6670\u6027&#xff0c;\u610f\u56fe\u6062\u590d\u672a\u7ecf\u8fc7\u9a8c\u8bc1\u5668\u8fc7\u6ee4&#xff0c;\u56e0\u6b64\u9a8c\u8bc1\u5668\u9009\u62e9\u5728\u6b64\u4e0d\u8d77\u4f5c\u7528\u3002<\/p>\n<p>\u8868 4&#xff1a;Terminal-Bench 2.1 \u4e0a\u6e90\u8f68\u8ff9 SFT \u548c\u610f\u56fe\u6062\u590d\u7684\u6bd4\u8f83 (%)<\/p>\n<table>\n<tr>\u53d8\u4f53\u6570\u636e\u5927\u5c0fClaude CodeTerminus2-XML\u5e73\u5747<\/tr>\n<tbody>\n<tr>\n<td>Qwen3.5-27B<\/td>\n<td>&#8211;<\/td>\n<td>47.8<\/td>\n<td>46.2<\/td>\n<td>47.0<\/td>\n<\/tr>\n<tr>\n<td>\u6e90\u8f68\u8ff9<\/td>\n<td>35.8k<\/td>\n<td>33.0<\/td>\n<td>40.3<\/td>\n<td>36.7<\/td>\n<\/tr>\n<tr>\n<td>\u610f\u56fe\u6062\u590d<\/td>\n<td>35.8k<\/td>\n<td>51.3<\/td>\n<td>52.9<\/td>\n<td>52.1<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u91cd\u65b0\u89e3\u51b3\u4f18\u4e8e\u6e90\u8f68\u8ff9 SFT\u3002\u6e90\u8f68\u8ff9\u548c\u610f\u56fe\u6062\u590d\u4f7f\u7528\u76f8\u540c\u7684 Qwen3.5 \u804a\u5929\u6a21\u677f\u548c\u5de5\u5177\u8c03\u7528\u6a21\u5f0f\u8fdb\u884c SFT\u3002\u6e90 SFT \u4fdd\u7559\u4e86\u539f\u59cb\u667a\u80fd\u4f53\u7684\u884c\u4e3a&#xff0c;\u800c\u610f\u56fe\u6062\u590d\u4f7f\u7528\u66f4\u5f3a\u7684\u6559\u5e08\u6765\u5728\u91cd\u5efa\u7684\u5de5\u4f5c\u7a7a\u95f4\u4e2d\u91cd\u65b0\u89e3\u51b3\u6062\u590d\u7684\u4efb\u52a1\u3002<\/p>\n<h4>6.2 \u667a\u80fd\u4f53\u8865\u5168\u7684\u5f71\u54cd<\/h4>\n<p>\u8fd9\u9879\u6d88\u878d\u7814\u7a76\u8be2\u95ee\u667a\u80fd\u4f53\u8865\u5168\u662f\u5426\u7269\u6709\u6240\u503c&#xff0c;\u8fd8\u662f\u4ec5\u786e\u5b9a\u6027\u56de\u653e\u5c31\u8db3\u591f\u4e86\u3002\u6211\u4eec\u901a\u8fc7\u5728\u4e0e\u610f\u56fe\u6062\u590d\u8bed\u6599\u5e93\u4e0a\u8bad\u7ec3\u6765\u9694\u79bb\u7b2c\u4e8c\u4e2a\u91cd\u5efa\u9636\u6bb5&#xff0c;\u8fd9\u4e9b\u73af\u5883\u4ec5\u4f7f\u7528\u56de\u653e\u91cd\u5efa&#xff0c;\u6ca1\u6709\u8865\u5168\u3002\u4ec5\u56de\u653e\u53d8\u4f53\u5728\u786e\u5b9a\u6027\u56de\u653e\u91cd\u5efa\u7684 35,809 \u4e2a\u6e90\u5b9e\u4f8b\u4e0a\u8bad\u7ec3&#xff0c;\u800c\u8865\u5168\u53d8\u4f53\u662f\u76f8\u540c\u6570\u91cf\u7684\u751f\u4ea7\u7ea7\u610f\u56fe\u6062\u590d\u8bed\u6599\u5e93\u3002\u4e24\u8005\u5171\u4eab\u6062\u590d\u7684\u67e5\u8be2\u548c\u6559\u5e08\u914d\u7f6e&#xff0c;\u4ec5\u5728\u73af\u5883\u91cd\u5efa\u6a21\u5f0f\u4e0a\u6709\u6240\u4e0d\u540c\u3002<\/p>\n<p>\u8868 5&#xff1a;\u667a\u80fd\u4f53\u8865\u5168\u5bf9\u610f\u56fe\u6062\u590d\u7684\u5f71\u54cd\u3002<\/p>\n<table>\n<tr>\u53d8\u4f53\u8bb0\u5f55\u6570Terminus2-XML<\/tr>\n<tbody>\n<tr>\n<td>Qwen3.5-27B<\/td>\n<td>&#8211;<\/td>\n<td>46.2<\/td>\n<\/tr>\n<tr>\n<td>\u4ec5\u56de\u653e<\/td>\n<td>35.8k<\/td>\n<td>48.7<\/td>\n<\/tr>\n<tr>\n<td>\u56de\u653e &#043; \u667a\u80fd\u4f53\u8865\u5168<\/td>\n<td>35.8k<\/td>\n<td>52.9<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u667a\u80fd\u4f53\u8865\u5168\u6709\u6240\u5e2e\u52a9\u3002\u5728\u76f8\u8fd1\u7684\u8bad\u7ec3\u91cf\u4e0b&#xff0c;\u5728\u8865\u5168\u73af\u5883\u4e0a\u8bad\u7ec3\u6bd4\u5728\u4ec5\u56de\u653e\u73af\u5883\u4e0a\u8bad\u7ec3\u9ad8\u51fa 4.2 \u4e2a\u767e\u5206\u70b9&#xff08;52.9 \u5bf9 48.7&#xff09;&#xff0c;\u5e76\u4e14\u8fd0\u884c\u66f4\u7a33\u5b9a&#xff08;\u00b11.4 \u5bf9 \u00b13.5&#xff09;\u3002\u8fd9\u4e0e\u8868 2 \u4e2d\u7684\u5145\u5206\u6027\u7ed3\u679c\u4e00\u81f4&#xff1a;\u4ec5\u56de\u653e\u4f7f\u5927\u591a\u6570\u7ec8\u7aef\u5de5\u4f5c\u7a7a\u95f4\u4efb\u52a1\u4e0d\u5145\u5206&#xff0c;\u800c\u8865\u5168\u6062\u590d\u4e86\u91cd\u65b0\u89e3\u51b3\u6240\u4f9d\u8d56\u7684\u6267\u884c\u4e0a\u4e0b\u6587\u3002<\/p>\n<p>\u4ec5\u56de\u653e\u76d1\u7763\u4ecd\u7136\u6bd4\u57fa\u7840\u6a21\u578b\u6709\u6240\u6539\u8fdb\u3002\u4ec5\u56de\u653e\u53d8\u4f53\u4ecd\u7136\u6bd4\u57fa\u7840\u6a21\u578b\u63d0\u9ad8\u4e86 2.5 \u4e2a\u767e\u5206\u70b9&#xff08;\u4ece 46.2 \u5230 48.7&#xff09;\u3002\u68c0\u67e5\u5176\u5c55\u5f00\u8fc7\u7a0b\u53d1\u73b0&#xff0c;\u5f53\u9762\u5bf9\u4e0d\u5b8c\u6574\u7684\u5de5\u4f5c\u7a7a\u95f4\u65f6&#xff0c;\u6559\u5e08\u901a\u5e38\u4f1a\u5728\u5904\u7406\u4efb\u52a1\u4e4b\u524d\u4fee\u590d\u7f3a\u5931\u7684\u6587\u4ef6\u6216\u8fdb\u884c\u8bbe\u7f6e\u3002\u8fd9\u4e9b\u8f68\u8ff9\u4ecd\u7136\u6709\u7528&#xff0c;\u4f46\u90e8\u5206\u76d1\u7763\u4fa7\u91cd\u4e8e\u4fee\u590d\u5de5\u4f5c\u7a7a\u95f4\u800c\u4e0d\u662f\u89e3\u51b3\u6062\u590d\u7684\u4efb\u52a1\u3002<\/p>\n<h4>6.3 \u9a8c\u8bc1\u5668\u8fc7\u6ee4\u7684\u4f5c\u7528<\/h4>\n<p>\u8fd9\u9879\u6d88\u878d\u7814\u7a76\u8be2\u95ee\u4e22\u5f03\u9a8c\u8bc1\u5931\u8d25\u8f68\u8ff9\u662f\u5426\u503c\u5f97\u79fb\u9664\u7684\u6570\u636e\u91cf\u3002\u5bf9\u4e8e\u6bcf\u79cd\u5408\u6210\u65b9\u6cd5&#xff0c;\u6211\u4eec\u6bd4\u8f83\u5728\u540c\u4e00\u542f\u52a8\u4efb\u52a1\u6c60\u4e2d&#xff0c;\u4f7f\u7528\u6240\u6709\u6559\u5e08\u8f68\u8ff9\u4e0e\u4ec5\u4f7f\u7528\u901a\u8fc7\u9a8c\u8bc1\u7684\u5b50\u96c6\u8fdb\u884c\u8bad\u7ec3\u7684\u6548\u679c\u3002<\/p>\n<p>\u8868 6&#xff1a;\u9a8c\u8bc1\u5668\u8fc7\u6ee4\u5bf9\u5355\u5de5\u4f5c\u7a7a\u95f4\u548c\u8de8\u5de5\u4f5c\u7a7a\u95f4\u7684\u5f71\u54cd\u3002<\/p>\n<table>\n<tr>\u53d8\u4f53\u8bb0\u5f55\u6570Terminus2-XML<\/tr>\n<tbody>\n<tr>\n<td>\u5355\u5de5\u4f5c\u7a7a\u95f4 w\/o \u9a8c\u8bc1\u5668<\/td>\n<td>35.1k<\/td>\n<td>56.0<\/td>\n<\/tr>\n<tr>\n<td>\u5355\u5de5\u4f5c\u7a7a\u95f4 w\/ \u9a8c\u8bc1\u5668<\/td>\n<td>25.4k<\/td>\n<td>56.4<\/td>\n<\/tr>\n<tr>\n<td>\u8de8\u5de5\u4f5c\u7a7a\u95f4 w\/o \u9a8c\u8bc1\u5668<\/td>\n<td>7.1k<\/td>\n<td>53.2<\/td>\n<\/tr>\n<tr>\n<td>\u8de8\u5de5\u4f5c\u7a7a\u95f4 w\/ \u9a8c\u8bc1\u5668<\/td>\n<td>3.5k<\/td>\n<td>55.4<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u9a8c\u8bc1\u5668\u8fc7\u6ee4\u5bf9\u4e8e\u66f4\u56f0\u96be\u7684\u4efb\u52a1\u66f4\u4e3a\u91cd\u8981\u3002\u5bf9\u4e8e\u6bcf\u79cd\u5408\u6210\u65b9\u6cd5&#xff0c;\u6211\u4eec\u6bd4\u8f83\u5728\u540c\u4e00\u542f\u52a8\u4efb\u52a1\u6c60\u4e2d&#xff0c;\u6240\u6709\u6559\u5e08\u8f68\u8ff9\u4e0e\u901a\u8fc7\u9a8c\u8bc1\u7684\u5b50\u96c6\u3002\u5728\u5355\u5de5\u4f5c\u7a7a\u95f4\u6570\u636e\u4e0a&#xff0c;\u901a\u8fc7\u9a8c\u8bc1\u7684\u5b50\u96c6\u8868\u73b0\u4e0e\u5b8c\u6574\u8bed\u6599\u5e93\u76f8\u4f3c&#xff08;56.4 \u5bf9 56.0&#xff09;&#xff0c;\u4e14\u8bb0\u5f55\u66f4\u5c11\u3002\u5728\u8de8\u5de5\u4f5c\u7a7a\u95f4\u6570\u636e\u4e0a&#xff0c;\u4fdd\u7559\u9a8c\u8bc1\u5931\u8d25\u7684\u8f68\u8ff9\u4f1a\u5c06\u6027\u80fd\u964d\u4f4e\u81f3 53.2&#xff0c;\u800c\u901a\u8fc7\u9a8c\u8bc1\u7684\u5b50\u96c6\u4ee5\u4e0d\u5230\u4e00\u534a\u7684\u6570\u636e\u8fbe\u5230\u4e86 55.4\u3002\u8fd9\u79cd\u6548\u679c\u5728\u591a\u8f6e\u5408\u6210\u4e2d\u4e5f\u5f88\u660e\u663e&#xff1a;\u79fb\u9664\u8f6e\u6b21\u7ea7\u9a8c\u8bc1\u76d1\u7763\u4f1a\u964d\u4f4e EvoCode-Bench \u7684\u4e24\u9879\u6307\u6807&#xff08;\u8868 9&#xff09;\u3002<\/p>\n<h4>6.4 \u5e7f\u5ea6\u6269\u5c55<\/h4>\n<p>\u8fd9\u9879\u6d88\u878d\u7814\u7a76\u8be2\u95ee\u8de8\u5de5\u4f5c\u7a7a\u95f4\u4efb\u52a1\u662f\u5426\u5728\u5355\u5de5\u4f5c\u7a7a\u95f4\u5408\u6210\u4e4b\u5916\u589e\u52a0\u4e86\u76d1\u7763\u3002\u8868 7 \u8bc4\u4f30\u4e86\u8de8\u5de5\u4f5c\u7a7a\u95f4\u76f8\u5bf9\u4e8e\u5355\u5de5\u4f5c\u7a7a\u95f4\u57fa\u7ebf\u7684\u8868\u73b0\u3002\u5355\u72ec\u7684\u8de8\u5de5\u4f5c\u7a7a\u95f4\u5728 Terminus2-XML \u4e0b\u8fbe\u5230 55.4&#xff0c;\u800c\u5c06\u5176\u6dfb\u52a0\u5230\u5355\u5de5\u4f5c\u7a7a\u95f4\u4f1a\u5c06\u6027\u80fd\u4ece 56.4 \u63d0\u5347\u81f3 58.4\u3002<\/p>\n<p>\u8868 7&#xff1a;Terminal-Bench 2.1 \u4e0a\u5355\u5de5\u4f5c\u7a7a\u95f4\u3001\u8de8\u5de5\u4f5c\u7a7a\u95f4\u53ca\u5176\u6df7\u5408\u7684\u7ed3\u679c (%)<\/p>\n<table>\n<tr>\u53d8\u4f53\u6570\u636e\u5927\u5c0fTerminus2-XML<\/tr>\n<tbody>\n<tr>\n<td>Qwen3.5-27B<\/td>\n<td>&#8211;<\/td>\n<td>46.2<\/td>\n<\/tr>\n<tr>\n<td>\u5355\u5de5\u4f5c\u7a7a\u95f4<\/td>\n<td>25.4k<\/td>\n<td>56.4<\/td>\n<\/tr>\n<tr>\n<td>\u8de8\u5de5\u4f5c\u7a7a\u95f4<\/td>\n<td>3.5k<\/td>\n<td>55.4<\/td>\n<\/tr>\n<tr>\n<td>\u5355\u5de5\u4f5c\u7a7a\u95f4 &#043; \u8de8\u5de5\u4f5c\u7a7a\u95f4<\/td>\n<td>28.9k<\/td>\n<td>58.4<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u8868 8&#xff1a;\u5355\u5de5\u4f5c\u7a7a\u95f4\u548c\u8de8\u5de5\u4f5c\u7a7a\u95f4\u5408\u6210\u7684\u8f68\u8ff9\u4e0e\u96be\u5ea6\u7279\u5f81\u3002<\/p>\n<table>\n<tr>\u53d8\u4f53\u8f6e\u6b21\u5de5\u5177\u8c03\u7528Token \/ \u8bb0\u5f55\u6559\u5e08 pass&#064;1<\/tr>\n<tbody>\n<tr>\n<td><\/td>\n<td>\u4e2d\u4f4d\u6570<\/td>\n<td>\u5747\u503c<\/td>\n<td>\u4e2d\u4f4d\u6570<\/td>\n<td>\u5747\u503c<\/td>\n<td>\u4e2d\u4f4d\u6570<\/td>\n<td>\u5747\u503c<\/td>\n<\/tr>\n<tr>\n<td>\u5355\u5de5\u4f5c\u7a7a\u95f4<\/td>\n<td>14<\/td>\n<td>15.1<\/td>\n<td>20<\/td>\n<td>21.2<\/td>\n<td>30.4k<\/td>\n<td>32.9k<\/td>\n<td>72.3%<\/td>\n<\/tr>\n<tr>\n<td>\u8de8\u5de5\u4f5c\u7a7a\u95f4<\/td>\n<td>23<\/td>\n<td>25.3<\/td>\n<td>38<\/td>\n<td>39.7<\/td>\n<td>46.5k<\/td>\n<td>51.9k<\/td>\n<td>49.2%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u8868 8 \u5bf9\u6bd4\u4e86\u540c\u4e00\u6559\u5e08\u4e0b\u7684\u4e24\u79cd\u5408\u6210\u53d8\u4f53\u3002\u8de8\u5de5\u4f5c\u7a7a\u95f4\u4efb\u52a1\u4ea7\u751f\u4e86\u663e\u8457\u66f4\u957f\u4e14\u66f4\u590d\u6742\u7684\u8f68\u8ff9&#xff1a;\u52a9\u7406\u8f6e\u6b21\u662f 1.6 \u500d&#xff0c;\u5de5\u5177\u8c03\u7528\u662f 1.9 \u500d&#xff0c;\u6bcf\u4e2a\u8bb0\u5f55\u7684\u4e2d\u4f4d\u6570 token \u662f 1.5 \u500d\u3002\u5b83\u4eec\u5bf9\u6559\u5e08\u6765\u8bf4\u4e5f\u66f4\u96be&#xff1a;pass&#064;1 \u4ece\u5355\u5de5\u4f5c\u7a7a\u95f4\u4efb\u52a1\u7684 72.3% \u4e0b\u964d\u5230 49.2%\u3002\u5c06\u9700\u6c42\u57fa\u4e8e\u7b2c\u4e8c\u4e2a\u4ee3\u7801\u5e93\u8feb\u4f7f\u667a\u80fd\u4f53\u5728\u7f16\u8f91\u4e4b\u524d\u9605\u8bfb\u5e76\u534f\u8c03\u4e24\u4e2a\u4ee3\u7801\u4ed3\u5e93&#xff0c;\u8fd9\u5ef6\u957f\u4e86\u8f68\u8ff9\u5e76\u63d0\u9ad8\u4e86\u4efb\u52a1\u96be\u5ea6&#xff0c;\u800c\u4e0d\u4ec5\u4ec5\u662f\u589e\u52a0\u6570\u636e\u91cf\u3002<\/p>\n<p>\u8de8\u5de5\u4f5c\u7a7a\u95f4\u5982\u4f55\u6269\u5c55\u5e7f\u5ea6\u3002\u00a0\u5355\u5de5\u4f5c\u7a7a\u95f4\u901a\u8fc7\u4ece\u6bcf\u4e2a\u6062\u590d\u7684\u73af\u5883\u4e2d\u5408\u6210\u4e00\u4e2a\u65b0\u4efb\u52a1\u6765\u63d0\u4f9b\u5de5\u4f5c\u7a7a\u95f4\u5185\u7684\u57fa\u7ebf\u3002\u8de8\u5de5\u4f5c\u7a7a\u95f4\u5219\u901a\u8fc7\u8981\u6c42\u5728\u53ef\u5199\u5de5\u4f5c\u7a7a\u95f4\u4e2d\u8fdb\u884c\u66f4\u6539&#xff0c;\u5e76\u5c06\u8fd9\u4e9b\u66f4\u6539\u57fa\u4e8e\u76f8\u5173\u7684\u53ea\u8bfb\u5de5\u4f5c\u7a7a\u95f4\u4e2d\u7684\u8bc1\u636e&#xff0c;\u4ece\u800c\u6269\u5c55\u4e86\u8fd9\u4e00\u8bbe\u7f6e\u3002\u8fd9\u66b4\u9732\u4e86\u5355\u5de5\u4f5c\u7a7a\u95f4\u4efb\u52a1\u4e2d\u4e0d\u5b58\u5728\u7684\u9886\u57df-\u64cd\u4f5c\u7ec4\u5408&#xff0c;\u5e76\u8981\u6c42\u667a\u80fd\u4f53\u8de8\u4ee3\u7801\u5e93\u534f\u8c03\u4fe1\u606f\u3002<\/p>\n<p>\u8de8\u5de5\u4f5c\u7a7a\u95f4\u6570\u636e\u5728\u4f55\u5904\u6709\u5e2e\u52a9\u3002\u00a0\u6df7\u5408\u5355\u5de5\u4f5c\u7a7a\u95f4\u548c\u8de8\u5de5\u4f5c\u7a7a\u95f4\u8bb0\u5f55\u5728 Terminus2-XML \u4e0b\u8fbe\u5230 58.4&#xff0c;\u76f8\u6bd4\u5355\u5de5\u4f5c\u7a7a\u95f4\u57fa\u7ebf (56.4) \u6709\u6240\u63d0\u9ad8\u3002\u56fe 6 \u6309\u4efb\u52a1\u7c7b\u522b\u7ec6\u5206\u4e86\u8fd9\u79cd\u6bd4\u8f83\u3002\u8f6f\u4ef6\u5de5\u7a0b\u662f\u6700\u5927\u7684\u7c7b\u522b&#xff0c;\u5305\u542b 26 \u4e2a\u4efb\u52a1&#xff0c;\u4ece 46.9 \u63d0\u9ad8\u5230 49.1&#xff0c;\u800c\u66f4\u5927\u7684\u63d0\u5347\u51fa\u73b0\u5728\u6a21\u578b\u8bad\u7ec3 (&#043;15.0) \u548c\u8c03\u8bd5 (&#043;10.0) \u4e2d\u3002\u6709\u4e09\u4e2a\u7c7b\u522b\u4e0b\u964d\u4e86 2.1 \u5230 2.5 \u4e2a\u767e\u5206\u70b9&#xff0c;\u8fd9\u5c5e\u4e8e\u6700\u591a\u516b\u4e2a\u4efb\u52a1\u7c7b\u522b\u7684\u8fd0\u884c\u95f4\u53d8\u5f02\u8303\u56f4\u5185\u3002\u5728\u8fd9\u4e2a\u6700\u5927\u7c7b\u522b\u4e2d\u7684\u63d0\u5347&#xff0c;\u52a0\u4e0a\u603b\u4f53\u6539\u8fdb&#xff0c;\u8868\u660e\u8de8\u5de5\u4f5c\u7a7a\u95f4\u6570\u636e\u5728\u5355\u5de5\u4f5c\u7a7a\u95f4\u5408\u6210\u4e4b\u5916\u589e\u52a0\u4e86\u6709\u7528\u7684\u76d1\u7763\u3002<\/p>\n<p class=\"img-center\"><img loading=\"lazy\" decoding=\"async\" alt=\"\" height=\"958\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260905232629-6a9ca52516873.png\" width=\"844\" \/><\/p>\n<p>\u56fe 6&#xff1a;\u5728 Terminus2-XML \u4e0b&#xff0c;\u5c06\u8de8\u5de5\u4f5c\u7a7a\u95f4\u6570\u636e\u6dfb\u52a0\u5230\u5355\u5de5\u4f5c\u7a7a\u95f4\u8bad\u7ec3\u4e2d\u65f6&#xff0c;Terminal-Bench 2.1 \u5404\u7c7b\u522b\u901a\u8fc7\u7387\u7684\u53d8\u5316\u3002\u7701\u7565\u4e86\u5355\u4efb\u52a1\u7c7b\u522b\u3002<\/p>\n<h4>6.5 \u591a\u8f6e\u6df1\u5ea6\u6269\u5c55<\/h4>\n<p>\u8fd9\u9879\u6d88\u878d\u7814\u7a76\u8be2\u95ee\u591a\u8f6e\u5ef6\u7eed\u6570\u636e\u662f\u5426\u80fd\u63d0\u9ad8\u5728\u6301\u4e45\u6027\u3001\u7d2f\u79ef\u6027\u4efb\u52a1\u4e0a\u7684\u6027\u80fd&#xff0c;\u4ee5\u53ca\u751f\u6210\u8fc7\u7a0b\u4e2d\u7684\u8f6e\u6b21\u7ea7\u9a8c\u8bc1\u662f\u5426\u91cd\u8981\u3002\u6211\u4eec\u5728 EvoCode-Bench v2 \u4e0a\u8fdb\u884c\u8bc4\u4f30\u3002<\/p>\n<p>\u591a\u8f6e\u6570\u636e\u63d0\u9ad8\u4e86\u6df1\u5ea6\u3002\u00a0\u5c06\u591a\u8f6e\u6570\u636e\u6dfb\u52a0\u5230\u5355\u5de5\u4f5c\u7a7a\u95f4\u4e2d&#xff0c;\u5c06 MT&#064;4 \u4ece 18.4 \u63d0\u9ad8\u5230 21.0&#xff0c;\u6848\u4f8b\u5206\u6570\u4ece 71.9 \u63d0\u9ad8\u5230 76.9\u3002\u8f83\u9ad8\u7684 MT&#064;4 \u8868\u660e\u6a21\u578b\u5728\u9996\u6b21\u5931\u8d25\u4e4b\u524d\u5b8c\u6210\u4e86\u66f4\u957f\u7684\u8bf7\u6c42\u5e8f\u5217&#xff0c;\u800c\u6848\u4f8b\u5206\u6570\u7684\u63d0\u9ad8\u8868\u660e\u5b83\u5728\u6574\u4e2a\u4f1a\u8bdd\u4e2d\u901a\u8fc7\u4e86\u66f4\u5927\u6bd4\u4f8b\u7684\u9a8c\u8bc1\u5668\u6848\u4f8b\u3002<\/p>\n<p>\u8f6e\u6b21\u7ea7\u9a8c\u8bc1\u6307\u5bfc\u6709\u6548\u7684\u5ef6\u7eed\u3002\u00a0\u5728\u5339\u914d\u7684\u5355\u5de5\u4f5c\u7a7a\u95f4 &#043; \u591a\u8f6e\u6bd4\u8f83\u4e2d&#xff0c;\u79fb\u9664\u8f6e\u6b21\u7ea7\u9a8c\u8bc1\u5668\u53cd\u9988\u4f1a\u4f7f MT&#064;4 \u964d\u4f4e 2.2 \u4e2a\u767e\u5206\u70b9&#xff0c;\u6848\u4f8b\u5206\u6570\u964d\u4f4e 3.7\u4e2a\u767e\u5206\u70b9&#xff08;\u8868 9&#xff09;\u3002\u8fd9\u8868\u660e&#xff0c;\u5728\u751f\u6210\u8fc7\u7a0b\u4e2d&#xff0c;\u5173\u4e8e\u4f1a\u8bdd\u6210\u529f\u6216\u5931\u8d25\u7684\u4fe1\u53f7\u6709\u52a9\u4e8e\u4ea7\u751f\u66f4\u6709\u6548\u7684\u591a\u8f6e\u6f14\u793a&#xff0c;\u8fd9\u6bd4\u7b80\u5355\u5730\u5c06\u72ec\u7acb\u7684\u5355\u8f6e\u8f68\u8ff9\u62fc\u63a5\u8d77\u6765\u66f4\u6709\u6548\u3002<\/p>\n<p>\u8868 9&#xff1a;EvoCode-Bench v2 \u4e0a\u591a\u8f6e\u5ef6\u7eed\u7684\u6d88\u878d\u7814\u7a76 (%)<\/p>\n<table>\n<tr>\u8bad\u7ec3\u6570\u636eMT&#064;4\u6848\u4f8b\u5206\u6570<\/tr>\n<tbody>\n<tr>\n<td>\u5355\u5de5\u4f5c\u7a7a\u95f4<\/td>\n<td>18.4<\/td>\n<td>71.9<\/td>\n<\/tr>\n<tr>\n<td>\u5355\u5de5\u4f5c\u7a7a\u95f4 &#043; \u591a\u8f6e (w\/o \u8f6e\u6b21\u7ea7\u9a8c\u8bc1)<\/td>\n<td>18.8<\/td>\n<td>73.2<\/td>\n<\/tr>\n<tr>\n<td>\u5355\u5de5\u4f5c\u7a7a\u95f4 &#043; \u591a\u8f6e (w\/ \u8f6e\u6b21\u7ea7\u9a8c\u8bc1)<\/td>\n<td>21.0<\/td>\n<td>76.9<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h4>6.6 \u73af\u5883\u6570\u91cf\u4e0e\u6bcf\u4e2a\u73af\u5883\u7684\u67e5\u8be2\u6570\u91cf<\/h4>\n<p>\u8fd9\u9879\u6d88\u878d\u7814\u7a76\u8be2\u95ee&#xff0c;\u5728\u56fa\u5b9a\u9884\u7b97\u4e0b&#xff0c;\u662f\u5426\u5e94\u4f18\u5148\u8003\u8651\u66f4\u591a\u73af\u5883\u8fd8\u662f\u6bcf\u4e2a\u73af\u5883\u7684\u66f4\u591a\u67e5\u8be2\u3002\u6211\u4eec\u56fa\u5b9a\u603b\u8bad\u7ec3\u8bb0\u5f55\u6570\u7ea6\u4e3a 25k&#xff0c;\u5e76\u6bd4\u8f83\u6e90\u81ea 25k \u4e2a\u4e0d\u540c\u73af\u5883&#xff08;\u6bcf\u4e2a\u73af\u5883\u4e00\u4e2a\u67e5\u8be2&#xff09;\u7684\u8bb0\u5f55\u4e0e\u6e90\u81ea 5k \u4e2a\u4e0d\u540c\u73af\u5883&#xff08;\u6bcf\u4e2a\u73af\u5883\u4e94\u4e2a\u67e5\u8be2&#xff09;\u7684\u8bb0\u5f55\u3002<\/p>\n<p>\u8868 10&#xff1a;\u73af\u5883\u6570\u91cf\u4e0e\u6bcf\u4e2a\u73af\u5883\u67e5\u8be2\u6570\u91cf\u7684\u6bd4\u8f83\u3002\u6240\u6709\u53d8\u4f53\u5747\u4f7f\u7528\u5355\u5de5\u4f5c\u7a7a\u95f4\u5408\u6210&#xff0c;\u56fa\u5b9a\u603b\u8bb0\u5f55\u6570\u7ea6\u4e3a 25k\u3002<\/p>\n<table>\n<tr>\u53d8\u4f53\u4e0d\u540c\u73af\u5883\u8bb0\u5f55\u6570Terminus2-XML<\/tr>\n<tbody>\n<tr>\n<td>1 \u4e2a\u67e5\u8be2 \/ \u73af\u5883<\/td>\n<td>25.4k<\/td>\n<td>25.4k<\/td>\n<td>56.4<\/td>\n<\/tr>\n<tr>\n<td>5 \u4e2a\u67e5\u8be2 \/ \u73af\u5883<\/td>\n<td>5.0k<\/td>\n<td>25.0k<\/td>\n<td>55.0<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u5728\u56fa\u5b9a\u8bb0\u5f55\u6570\u4e0b&#xff0c;\u4ece\u66f4\u591a\u73af\u5883\u4e2d\u53d6\u6837\u8868\u73b0\u66f4\u597d\u3002\u8868 10 \u663e\u793a&#xff0c;\u4ece 25k \u4e2a\u4e0d\u540c\u73af\u5883\u4e2d\u5404\u53d6\u4e00\u4e2a\u67e5\u8be2 (56.4) \u4f18\u4e8e\u4ece 5k \u4e2a\u73af\u5883\u4e2d\u5404\u53d6\u4e94\u4e2a\u67e5\u8be2 (55.0)\u3002\u5728\u56fa\u5b9a\u9884\u7b97\u4e0b&#xff0c;\u4f18\u5148\u8003\u8651\u73af\u5883\u591a\u6837\u6027\u6bd4\u589e\u52a0\u6bcf\u4e2a\u73af\u5883\u7684\u67e5\u8be2\u6570\u66f4\u6709\u6548\u3002<\/p>\n<h4>6.7 \u6cdb\u5316\u5230 SWE \u57df<\/h4>\n<p>\u4e3a\u4e86\u6d4b\u8bd5\u8be5\u65b9\u6cd5\u662f\u5426\u9002\u7528\u4e8e\u7ec8\u7aef\u5de5\u4f5c\u7a7a\u95f4\u4e4b\u5916&#xff0c;\u6211\u4eec\u5728\u4ece SWE \u6c60\u91cd\u5efa\u7684\u73af\u5883\u4e0a\u8fd0\u884c\u610f\u56fe\u6062\u590d&#xff0c;\u5e76\u5728 Terminal-Bench 2.1 \u4e0a\u8bc4\u4f30\u6240\u5f97\u6a21\u578b&#xff0c;\u8be5\u57fa\u51c6\u5305\u542b\u8f6f\u4ef6\u5de5\u7a0b\u4efb\u52a1\u3002<\/p>\n<p>\u8868 11&#xff1a;\u5728 Terminal-Bench 2.1 \u4e0a&#xff0c;\u5728 SWE \u91cd\u5efa\u73af\u5883\u4e0a\u8fdb\u884c\u610f\u56fe\u6062\u590d\u7684\u7ed3\u679c\u3002<\/p>\n<table>\n<tr>\u8bad\u7ec3\u6570\u636e\u8bb0\u5f55\u6570Terminus2-XML<\/tr>\n<tbody>\n<tr>\n<td>Qwen3.5-27B<\/td>\n<td>&#8211;<\/td>\n<td>46.2<\/td>\n<\/tr>\n<tr>\n<td>\u7ec8\u7aef\u6c60\u610f\u56fe\u6062\u590d<\/td>\n<td>35.8k<\/td>\n<td>52.9<\/td>\n<\/tr>\n<tr>\n<td>SWE \u6c60\u610f\u56fe\u6062\u590d<\/td>\n<td>22.3k<\/td>\n<td>48.6<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>SWE \u91cd\u5efa\u5728\u5355\u72ec\u4f7f\u7528\u65f6\u63d0\u9ad8\u4e86\u6027\u80fd\u3002\u5728 SWE \u6062\u590d\u7684\u73af\u5883\u4e0a\u8bad\u7ec3\u8fbe\u5230 48.6&#xff0c;\u6bd4\u57fa\u7840\u6a21\u578b\u9ad8 2.4 \u4e2a\u767e\u5206\u70b9&#xff08;\u8868 11&#xff09;\u3002\u7136\u800c&#xff0c;\u5b83\u843d\u540e\u4e8e\u7ec8\u7aef\u6c60\u7248\u672c&#xff08;52.9&#xff09;&#xff0c;\u8fd9\u53ef\u80fd\u662f\u7531\u4e8e SWE \u91cd\u5efa\u7684\u53ef\u7528\u5de5\u4f5c\u7a7a\u95f4\u8f83\u5c11&#xff08;\u00a74.2 \u8868 2&#xff09;\u3002\u6211\u4eec\u5c06\u8fd9\u4e24\u4e2a\u6c60\u7684\u6df7\u5408\u7559\u7ed9\u672a\u6765\u7684\u5de5\u4f5c\u3002<\/p>\n<h3>7 \u8ba8\u8bba\u4e0e\u5c40\u9650\u6027<\/h3>\n<p>\u4f7f\u7528\u66f4\u4e30\u5bcc\u7684\u6e90\u8f68\u8ff9\u8fdb\u884c\u6269\u5c55\u3002\u00a0\u867d\u7136 Terminal-Universe \u5728\u5e94\u7528\u4e8e\u516c\u5171\u8f68\u8ff9\u8bed\u6599\u5e93\u65f6\u663e\u793a\u51fa\u5f3a\u5927\u7684\u6709\u6548\u6027&#xff0c;\u4f46\u6211\u4eec\u5bf9\u66f4\u9ad8\u8d28\u91cf\u8f68\u8ff9\u7684\u5b9e\u8bc1\u89c2\u5bdf\u8868\u660e&#xff0c;\u6e90\u8f68\u8ff9\u7684\u590d\u6742\u5ea6\u4e0e\u5728\u91cd\u5efa\u73af\u5883\u4e0a\u5408\u6210\u7684\u7ed3\u679c\u5c55\u5f00\u7684\u590d\u6742\u5ea6\u4e4b\u95f4\u5b58\u5728\u6b63\u76f8\u5173\u3002\u5177\u6709\u66f4\u4e30\u5bcc\u64cd\u4f5c\u52a8\u6001&#xff08;\u4f8b\u5982\u591a\u6587\u4ef6\u64cd\u4f5c\u548c\u957f\u5de5\u5177\u4f7f\u7528\u94fe&#xff09;\u7684\u8f68\u8ff9\u66b4\u9732\u4e86\u66f4\u5e7f\u6cdb\u7684\u5de5\u4f5c\u7a7a\u95f4\u72b6\u6001&#xff0c;\u8fd9\u53cd\u8fc7\u6765\u652f\u6301\u66f4\u591a\u6837\u5316\u548c\u8981\u6c42\u66f4\u9ad8\u7684\u5408\u6210\u4efb\u52a1\u3002\u73af\u5883\u590d\u6742\u5ea6\u4e5f\u53ef\u4ee5\u901a\u8fc7\u805a\u5408\u6765\u63d0\u9ad8&#xff1a;\u5f53\u591a\u4e2a\u4f1a\u8bdd\u5728\u540c\u4e00\u5de5\u4f5c\u7a7a\u95f4\u4e0a\u64cd\u4f5c\u65f6&#xff0c;\u5c06\u5b83\u4eec\u4e00\u8d77\u56de\u653e\u4f1a\u5c06\u6bcf\u4e2a\u4f1a\u8bdd\u66b4\u9732\u7684\u72b6\u6001\u5408\u5e76\u5230\u5355\u4e2a\u91cd\u5efa\u4e2d&#xff0c;\u4ea7\u751f\u6bd4\u4efb\u4f55\u5355\u4e2a\u8f68\u8ff9\u90fd\u80fd\u6062\u590d\u7684\u66f4\u590d\u6742\u7684\u73af\u5883\u3002\u603b\u4e4b&#xff0c;\u8fd9\u4e9b\u89c2\u5bdf\u8868\u660e&#xff0c;\u968f\u7740\u667a\u80fd\u4f53\u80fd\u529b\u7684\u8fdb\u6b65\u4f7f\u590d\u6742\u8f68\u8ff9\u8d8a\u6765\u8d8a\u5bb9\u6613\u83b7\u5f97&#xff0c;\u8be5\u6846\u67b6\u81ea\u7136\u4f1a\u6269\u5c55\u3002<\/p>\n<p>\u5c40\u9650\u6027\u3002\u00a0Terminal-Universe \u4e5f\u5b58\u5728\u4e00\u4e9b\u5c40\u9650\u6027\u3002\u9996\u5148&#xff0c;\u6211\u4eec\u4e3a\u6240\u6709\u5de5\u4f5c\u7a7a\u95f4\u4f7f\u7528\u6807\u51c6\u7684 Ubuntu 24.04 \u5bb9\u5668&#xff0c;\u800c\u4e0d\u662f\u901a\u8fc7\u50cf SWE-Factory (Guo et al., 2026) \u6216 RepoLaunch (Li et al., 2026a) \u8fd9\u6837\u7684\u81ea\u52a8\u5316\u6d41\u6c34\u7ebf\u6784\u5efa\u5b9a\u5236\u7684\u3001\u7279\u5b9a\u4e8e\u4ee3\u7801\u4ed3\u5e93\u7684\u73af\u5883\u3002\u56e0\u6b64&#xff0c;\u9700\u8981\u4e13\u95e8\u7cfb\u7edf\u4f9d\u8d56\u9879\u6216\u590d\u6742\u7f16\u8bd1\u6b65\u9aa4\u7684\u8fb9\u7f18\u60c5\u51b5\u53ef\u80fd\u4fdd\u771f\u5ea6\u8f83\u4f4e\u3002\u5176\u6b21&#xff0c;\u6211\u4eec\u91cd\u5efa\u5de5\u4f5c\u7a7a\u95f4\u7684\u9886\u57df\u3001\u8bed\u8a00\u548c\u5de5\u5177\u94fe\u5206\u5e03\u4ece\u6839\u672c\u4e0a\u53d7\u9650\u4e8e\u6536\u96c6\u5230\u7684\u6e90\u8f68\u8ff9\u7684\u8986\u76d6\u8303\u56f4\u3002\u8d85\u8d8a\u8fd9\u4e9b\u5206\u5e03\u8fdb\u884c\u6269\u5c55\u4ecd\u7136\u662f\u672a\u6765\u5de5\u4f5c\u7684\u91cd\u8981\u65b9\u5411\u3002\u7b2c\u4e09&#xff0c;\u5355\u4e2a\u6559\u5e08\u751f\u6210\u4efb\u52a1\u3001\u89e3\u51b3\u65b9\u6848\u548c\u9a8c\u8bc1\u5668\u3002\u56e0\u6b64&#xff0c;\u5176\u80fd\u529b\u5dee\u8ddd\u53ef\u80fd\u9650\u5236\u4efb\u52a1\u8986\u76d6\u8303\u56f4&#xff0c;\u5e76\u4e14\u5176\u5728\u89e3\u51b3\u65b9\u6848\u4e2d\u7684\u9519\u8bef\u4e5f\u53ef\u80fd\u88ab\u5176\u6d4b\u8bd5\u9057\u6f0f\u3002\u672a\u6765\u7684\u5de5\u4f5c\u53ef\u4ee5\u4f7f\u7528\u591a\u4e2a\u6559\u5e08\u548c\u4e00\u4e2a\u72ec\u7acb\u7684\u6a21\u578b\u8fdb\u884c\u9a8c\u8bc1\u5668\u6784\u5efa\u3002<\/p>\n<h3>8 \u7ed3\u8bba<\/h3>\n<p>\u5728\u8fd9\u9879\u5de5\u4f5c\u4e2d&#xff0c;\u6211\u4eec\u63d0\u51fa\u4e86 Terminal-Universe&#xff0c;\u4e00\u4e2a\u57fa\u4e8e\u8f68\u8ff9\u7684\u73af\u5883\u548c\u4efb\u52a1\u5408\u6210\u6846\u67b6&#xff0c;\u5b83\u5c06\u8bb0\u5f55\u7684\u667a\u80fd\u4f53\u8f68\u8ff9\u4ece\u56fa\u5b9a\u7684\u6f14\u793a\u91cd\u65b0\u5b9a\u4e49\u4e3a\u53ef\u6062\u590d\u3001\u53ef\u91cd\u7528\u7684\u6267\u884c\u73af\u5883\u3002\u901a\u8fc7\u786e\u5b9a\u6027\u56de\u653e\u548c\u667a\u80fd\u4f53\u8865\u5168\u7684\u4e24\u9636\u6bb5\u8fc7\u7a0b&#xff0c;Terminal-Universe \u6062\u590d\u4e86\u6f5c\u5728\u7684\u5de5\u4f5c\u7a7a\u95f4\u72b6\u6001&#xff0c;\u5728\u6bcf\u4e2a\u73af\u5883\u4e2d\u5408\u6210\u65b0\u4efb\u52a1&#xff0c;\u5e76\u8de8\u5de5\u4f5c\u7a7a\u95f4&#xff08;\u5e7f\u5ea6&#xff09;\u548c\u8de8\u8f6e\u6b21&#xff08;\u6df1\u5ea6&#xff09;\u6269\u5c55\u5b83\u4eec\u3002\u5b9e\u8bc1\u8bc4\u4f30\u8868\u660e&#xff0c;\u91cd\u65b0\u89e3\u51b3\u6062\u590d\u7684\u4efb\u52a1\u4f18\u4e8e\u5728\u6e90\u8f68\u8ff9\u4e0a\u8fdb\u884c SFT&#xff0c;\u800c\u7ecf\u8fc7\u9a8c\u8bc1\u5668\u9009\u62e9\u7684\u5355\u5de5\u4f5c\u7a7a\u95f4\u3001\u8de8\u5de5\u4f5c\u7a7a\u95f4\u548c\u591a\u8f6e\u8f68\u8ff9\u540c\u65f6\u63d0\u9ad8\u4e86\u5355\u8f6e\u548c\u591a\u8f6e\u7ec8\u7aef\u667a\u80fd\u4f53\u7684\u6027\u80fd\u3002\u603b\u4f53\u800c\u8a00&#xff0c;\u6211\u4eec\u7684\u7814\u7a76\u7ed3\u679c\u8868\u660e&#xff0c;\u8bb0\u5f55\u7684\u8f68\u8ff9\u53ef\u4ee5\u6709\u6548\u5730\u91cd\u65b0\u7528\u4e8e\u4ea4\u4e92\u5f0f\u6267\u884c\u73af\u5883&#xff0c;\u4e3a\u5408\u6210\u57fa\u4e8e\u771f\u5b9e\u573a\u666f\u7684\u667a\u80fd\u4f53\u8bad\u7ec3\u6570\u636e\u63d0\u4f9b\u4e86\u4e00\u79cd\u53ef\u6269\u5c55\u7684\u8303\u5f0f&#xff0c;\u800c\u65e0\u9700\u4ece\u5934\u6784\u5efa\u73af\u5883\u3002<\/p>\n<h3>A \u6570\u636e\u6e90\u4e0e\u9009\u62e9<\/h3>\n<p>\u8868 12 \u603b\u7ed3\u4e86\u91cd\u5efa\u524d\u540e\u7684\u6bcf\u4e2a\u5305\u542b\u6e90\u3002\u4e00\u6761\u8f68\u8ff9\u80fd\u5426\u88ab\u91cd\u5efa\u53d6\u51b3\u4e8e\u5176\u66b4\u9732\u7684\u6587\u4ef6\u8bc1\u636e&#xff1a;\u591a\u6587\u4ef6\u7f16\u8f91\u8f68\u8ff9\u63ed\u793a\u4e86\u8db3\u591f\u7684\u9879\u76ee\u4e0a\u4e0b\u6587\u4ee5\u4f9b\u91cd\u5efa&#xff0c;\u800c\u4ee5\u547d\u4ee4\u4e3a\u4e3b\u7684 CLI \u8f68\u8ff9\u901a\u5e38\u66b4\u9732\u5f88\u5c11\u7684\u6587\u4ef6\u5185\u5bb9\u3002\u6211\u4eec\u6392\u9664\u4e86\u6e90\u81ea Terminal-Bench \u7684\u8bed\u6599\u5e93\u4ee5\u53ca\u4ee5\u53ea\u8bfb\u64cd\u4f5c\u6216\u4e0d\u652f\u6301\u7684\u64cd\u4f5c\u683c\u5f0f\u4e3a\u4e3b\u7684\u6e90\u3002<\/p>\n<p>SWE \u53bb\u91cd\u4e0e\u53bb\u6c61\u3002\u00a0\u516c\u5171 SWE \u8bed\u6599\u5e93\u901a\u5e38\u5305\u542b\u540c\u4e00\u4efb\u52a1\u7684\u591a\u4e2a\u5c55\u5f00\u6216\u4e0d\u540c\u683c\u5f0f\u3002\u6211\u4eec\u5728\u6bcf\u4e2a\u6e90\u5185\u6309\u4ee3\u7801\u4ed3\u5e93\u3001\u57fa\u7840\u63d0\u4ea4\u548c\u95ee\u9898\u9648\u8ff0\u8fdb\u884c\u53bb\u91cd&#xff0c;\u9009\u62e9\u5176\u56de\u653e\u66b4\u9732\u6700\u591a\u5de5\u4f5c\u7a7a\u95f4\u5185\u5bb9\u7684\u8f68\u8ff9\u3002\u7136\u540e&#xff0c;\u6211\u4eec\u79fb\u9664\u6765\u81ea SWE-bench Verified \u4ee3\u7801\u4ed3\u5e93 (Jimenez et al., 2024) \u7684\u5b9e\u4f8b\u4ee5\u53ca\u65e0\u6cd5\u94fe\u63a5\u5230\u6e90\u8f68\u8ff9\u7684\u8bb0\u5f55\u3002\u8868 12 \u7ed9\u51fa\u4e86\u6700\u7ec8\u7684\u6bcf\u4e2a\u6e90\u8ba1\u6570\u3002<\/p>\n<p>\u8868 12&#xff1a;\u6e90\u8bed\u6599\u5e93\u548c\u91cd\u5efa\u73af\u5883\u3002\u8f68\u8ff9\u5217\u7edf\u8ba1\u4ece\u6bcf\u4e2a\u8bed\u6599\u5e93\u4e2d\u62bd\u53d6\u7684\u6e90\u8f68\u8ff9\u6570\u91cf\u3002<\/p>\n<table>\n<tr>\u6e90\u8bed\u6599\u5e93\u6c60\u8bb8\u53ef\u8bc1\u8f68\u8ff9\u6570\u73af\u5883\u6570<\/tr>\n<tbody>\n<tr>\n<td>SWE-rebench (Badertdinov et al., 2025)<\/td>\n<td>SWE<\/td>\n<td>CC-BY-4.0<\/td>\n<td>67,074<\/td>\n<td>6,118<\/td>\n<\/tr>\n<tr>\n<td>SWE-smith (Yang et al., 2025)<\/td>\n<td>SWE<\/td>\n<td>MIT<\/td>\n<td>95,851<\/td>\n<td>8,476<\/td>\n<\/tr>\n<tr>\n<td>CoderForge (Ariyak et al., 2026)<\/td>\n<td>SWE<\/td>\n<td>Apache-2.0<\/td>\n<td>32,964<\/td>\n<td>3,978<\/td>\n<\/tr>\n<tr>\n<td>SWE-Gym (Pan et al., 2024)<\/td>\n<td>SWE<\/td>\n<td>MIT<\/td>\n<td>4,152<\/td>\n<td>929<\/td>\n<\/tr>\n<tr>\n<td>LFM2-Terminal (gyung, 2026)<\/td>\n<td>\u7ec8\u7aef<\/td>\n<td>CC-BY-4.0<\/td>\n<td>139,841<\/td>\n<td>46,037<\/td>\n<\/tr>\n<tr>\n<td>LiteCoder-Terminal (Peng et al., 2026a)<\/td>\n<td>\u7ec8\u7aef<\/td>\n<td>MIT<\/td>\n<td>19,711<\/td>\n<td>2,725<\/td>\n<\/tr>\n<tr>\n<td>\u603b\u8ba1<\/td>\n<td><\/td>\n<td><\/td>\n<td>359,593<\/td>\n<td>68,263<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u8868 13&#xff1a;\u667a\u80fd\u4f53\u8865\u5168\u524d\u540e\u7684\u5de5\u4f5c\u7a7a\u95f4\u590d\u6742\u5ea6\u3002\u5355\u5143\u683c\u62a5\u544a\u4e2d\u4f4d\u6570 \/ \u5747\u503c\u3002<\/p>\n<table>\n<tr>\u7ec8\u7aef\u6c60SWE \u6c60<\/tr>\n<tbody>\n<tr>\n<td><\/td>\n<td>\u56de\u653e\u540e (E\u2080)<\/td>\n<td>\u8865\u5168\u540e (E)<\/td>\n<td>\u56de\u653e\u540e (E\u2080)<\/td>\n<td>\u8865\u5168\u540e (E)<\/td>\n<\/tr>\n<tr>\n<td>\u6bcf\u4e2a\u5de5\u4f5c\u7a7a\u95f4\u6587\u4ef6\u6570<\/td>\n<td>2 \/ 2.9<\/td>\n<td>13 \/ 22.4<\/td>\n<td>5 \/ 5.6<\/td>\n<td>21 \/ 37.9<\/td>\n<\/tr>\n<tr>\n<td>\u6587\u672c\u884c\u6570&#xff08;\u6240\u6709\u6587\u4ef6&#xff09;<\/td>\n<td>60 \/ 905<\/td>\n<td>539 \/ 5,761<\/td>\n<td>537 \/ 595<\/td>\n<td>1,854 \/ 6,622<\/td>\n<\/tr>\n<tr>\n<td>\u4ee3\u7801\u884c\u6570&#xff08;\u6e90\u6587\u4ef6&#xff09;<\/td>\n<td>0 \/ 433<\/td>\n<td>316 \/ 503<\/td>\n<td>421 \/ 487<\/td>\n<td>1,643 \/ 6,302<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p class=\"img-center\"><img loading=\"lazy\" decoding=\"async\" alt=\"\" height=\"812\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260905232629-6a9ca52551797.png\" width=\"1684\" \/><\/p>\n<p>\u56fe 7 \u603b\u7ed3\u4e86\u6838\u5fc3\u91cd\u5efa\u548c\u5145\u5206\u6027\u9636\u6bb5\u3002<\/p>\n<h4>B.1 \u91cd\u5efa\u4e0e\u8865\u5168<\/h4>\n<p>\u6e90\u8bed\u6599\u5e93\u4ee5\u4e0d\u540c\u65b9\u5f0f\u7f16\u7801\u6587\u4ef6\u64cd\u4f5c&#xff0c;\u56e0\u6b64\u6211\u4eec\u5c06\u8bfb\u53d6\u3001\u5199\u5165\u548c\u7f16\u8f91\u64cd\u4f5c\u89c4\u8303\u5316\u4e3a\u6709\u5e8f\u4e8b\u4ef6\u6d41\u3002\u6211\u4eec\u4fdd\u7559\u53ea\u8bfb\u6587\u4ef6\u7684\u89c2\u5bdf\u5185\u5bb9&#xff0c;\u5bf9\u4e8e\u4fee\u6539\u8fc7\u7684\u6587\u4ef6&#xff0c;\u4fdd\u7559\u9996\u6b21\u66f4\u6539\u524d\u53ef\u89c1\u7684\u5185\u5bb9\u3002\u8fd9\u4e9b\u5185\u5bb9\u53ef\u80fd\u662f\u90e8\u5206\u7684\u6216\u88ab\u622a\u65ad\u7684\u3002\u667a\u80fd\u4f53\u521b\u5efa\u7684\u6587\u4ef6\u548c\u6240\u6709\u540e\u7eed\u66f4\u6539\u90fd\u4ece\u56de\u653e\u7684\u5de5\u4f5c\u7a7a\u95f4\u4e2d\u6392\u9664\u3002<\/p>\n<p>\u8865\u5168\u667a\u80fd\u4f53\u63a5\u6536\u6062\u590d\u7684\u8bf7\u6c42\u3001\u90e8\u5206\u5de5\u4f5c\u7a7a\u95f4\u548c\u6587\u4ef6\u6e05\u5355\u3002\u5b83\u901a\u8fc7\u521b\u5efa\u6587\u4ef6\u3001\u8865\u5168\u90e8\u5206\u6587\u4ef6\u4ee5\u53ca\u914d\u7f6e\u73b0\u6709\u9879\u76ee\u6240\u9700\u7684\u4f9d\u8d56\u9879\u6765\u586b\u5145\u7f3a\u5931\u7684\u9879\u76ee\u4e0a\u4e0b\u6587\u3002\u5b83\u4e0d\u5f97\u5b9e\u73b0\u8bf7\u6c42\u7684\u4efb\u52a1\u6216\u900f\u9732\u5e94\u5728\u4f55\u5904\u8fdb\u884c\u89e3\u51b3\u65b9\u6848\u3002<\/p>\n<p>\u6253\u5305\u3002\u00a0\u91cd\u5efa\u7684\u9879\u76ee\u653e\u7f6e\u5728 \/app \u4e0b\u3002\u5728\u5176\u4ed6\u6587\u4ef6\u7cfb\u7edf\u4f4d\u7f6e\u8bbf\u95ee\u7684\u6587\u4ef6\u5219\u5355\u72ec\u5b58\u50a8\u3002<\/p>\n<p>\u63d0\u793a&#xff1a;\u73af\u5883\u8865\u5168<\/p>\n<p>[\u63d0\u793a\u8bcd\u5185\u5bb9&#xff0c;\u6b64\u5904\u7701\u7565\u4ee5\u8282\u7701\u7a7a\u95f4]<\/p>\n<p>\u8865\u5168\u8d28\u91cf\u3002\u00a0\u5bf9 30 \u4e2a\u968f\u673a\u62bd\u6837\u7684\u7ec8\u7aef\u8865\u5168\u8fdb\u884c\u4eba\u5de5\u68c0\u67e5\u53d1\u73b0&#xff0c;\u5176\u4e2d 22 \u4e2a\u4ec5\u5305\u542b\u4efb\u52a1\u76f8\u5173\u7684\u652f\u6301\u6587\u4ef6&#xff0c;\u800c 8 \u4e2a\u5f15\u5165\u4e86\u5927\u91cf\u975e\u652f\u6301\u4efb\u52a1\u6240\u9700\u7684\u6587\u4ef6\u6216\u4ee3\u7801\u3002\u5728 30 \u4e2a\u8865\u5168\u7684\u5de5\u4f5c\u7a7a\u95f4\u4e2d\u672a\u53d1\u73b0\u4efb\u52a1\u89e3\u51b3\u65b9\u6848\u3002<\/p>\n<h4>B.2 \u5de5\u4f5c\u7a7a\u95f4\u5145\u5206\u6027\u8bc4\u4f30<\/h4>\n<p>\u8bc4\u4f30\u8303\u56f4\u3002\u00a0\u6211\u4eec\u8bc4\u4f30\u6bcf\u4e2a\u5b58\u6d3b\u4e0b\u6765\u901a\u8fc7\u53bb\u6c61\u548c\u4ee3\u7801\u4ed3\u5e93\u7ea7\u53bb\u91cd\u7684\u7ec8\u7aef\u91cd\u5efa\u3002\u5bf9\u4e8e SWE&#xff0c;\u6211\u4eec\u4e3a\u6bcf\u4e2a\u4ee3\u7801\u4ed3\u5e93\u8bc4\u4f30\u4e00\u4e2a\u4ee3\u8868\u6027\u91cd\u5efa\u53ca\u5176\u610f\u56fe\u6062\u590d\u4efb\u52a1\u3002\u667a\u80fd\u4f53\u8bc4\u5224\u5458\u4f7f\u7528\u53ea\u8bfb shell \u548c\u6587\u4ef6\u5de5\u5177\u68c0\u67e5\u6bcf\u4e2a\u5de5\u4f5c\u7a7a\u95f4\u3002<\/p>\n<p>\u8bc4\u5206\u3002\u00a0\u5f53\u5de5\u4f5c\u7a7a\u95f4\u7684\u6e90\u4ee3\u7801\u3001\u914d\u7f6e\u3001\u6570\u636e\u548c\u7ed3\u6784\u4e3a\u6709\u80fd\u529b\u7684\u667a\u80fd\u4f53\u63d0\u4f9b\u4e86\u5904\u7406\u8be5\u4efb\u52a1\u6240\u9700\u7684\u8db3\u591f\u4e0a\u4e0b\u6587\u65f6&#xff0c;\u8be5\u5de5\u4f5c\u7a7a\u95f4\u88ab\u8ba4\u4e3a\u662f\u4efb\u52a1\u5145\u5206\u7684\u3002\u4f9d\u8d56\u9879\u3001\u7f13\u5b58\u3001\u751f\u6210\u8f93\u51fa\u548c\u53ef\u4ee5\u91cd\u65b0\u521b\u5efa\u7684\u53ef\u9009\u6587\u6863\u4e0d\u662f\u5fc5\u9700\u7684\u3002\u8bc4\u5224\u5458\u8bc4\u4f30\u53ef\u7528\u7684\u9879\u76ee\u4e0a\u4e0b\u6587&#xff0c;\u800c\u4e0d\u662f\u6784\u5efa\u6210\u529f\u3002<\/p>\n<p>\u8868 14 \u7ed9\u51fa\u4e86\u76f8\u5e94\u7684\u6e90\u7ea7\u522b\u7ec6\u5206\u3002<\/p>\n<p>\u8868 14&#xff1a;\u667a\u80fd\u4f53\u8865\u5168\u540e&#xff0c;\u5728\u6062\u590d\u610f\u56fe\u4e0b\u5de5\u4f5c\u7a7a\u95f4\u5145\u5206\u6027\u7684\u6e90\u7ec6\u5206\u3002<\/p>\n<table>\n<tr>\u6e90n\u5145\u5206\u5145\u5206\u7387 (%)<\/tr>\n<tbody>\n<tr>\n<td>\u7ec8\u7aef\u6c60<\/td>\n<td><\/td>\n<td><\/td>\n<td><\/td>\n<\/tr>\n<tr>\n<td>LFM2-Terminal<\/td>\n<td>36,204<\/td>\n<td>34,467<\/td>\n<td>95.2<\/td>\n<\/tr>\n<tr>\n<td>LiteCoder-Terminal<\/td>\n<td>2,090<\/td>\n<td>1,342<\/td>\n<td>64.2<\/td>\n<\/tr>\n<tr>\n<td>SWE \u6c60<\/td>\n<td><\/td>\n<td><\/td>\n<td><\/td>\n<\/tr>\n<tr>\n<td>SWE-rebench<\/td>\n<td>1,734<\/td>\n<td>1,350<\/td>\n<td>77.9<\/td>\n<\/tr>\n<tr>\n<td>SWE-smith<\/td>\n<td>114<\/td>\n<td>85<\/td>\n<td>74.6<\/td>\n<\/tr>\n<tr>\n<td>CoderForge<\/td>\n<td>47<\/td>\n<td>27<\/td>\n<td>57.4<\/td>\n<\/tr>\n<tr>\n<td>SWE-Gym<\/td>\n<td>5<\/td>\n<td>2<\/td>\n<td>40.0<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u5145\u5206\u7387\u53cd\u6620\u4e86\u8f68\u8ff9\u66b4\u9732\u4e86\u591a\u5c11\u9879\u76ee\u4e0a\u4e0b\u6587\u3002\u667a\u80fd\u4f53\u8865\u5168\u53ef\u4ee5\u5728\u89c2\u5bdf\u5230\u7684\u8bc1\u636e\u5468\u56f4\u6dfb\u52a0\u5e38\u89c1\u7684\u9879\u76ee\u6587\u4ef6&#xff0c;\u4f46\u65e0\u6cd5\u6062\u590d\u672a\u89c1\u8fc7\u7684\u9879\u76ee\u7279\u5b9a\u5185\u5bb9\u3002\u67d0\u4e9b SWE \u5b50\u96c6\u592a\u5c0f&#xff0c;\u65e0\u6cd5\u8fdb\u884c\u7ec6\u7c92\u5ea6\u6bd4\u8f83\u3002<\/p>\n<h4>B.3 \u5931\u8d25\u5206\u6790<\/h4>\n<p>\u5bf9\u4e8e\u4efb\u52a1\u4e0d\u5145\u5206\u7684\u5de5\u4f5c\u7a7a\u95f4&#xff0c;\u8bc4\u5224\u5458\u4f1a\u6307\u51fa\u7f3a\u5931\u6216\u622a\u65ad\u7684\u6e90\u6587\u4ef6\u3001\u4efb\u52a1\u811a\u672c\u3001\u8f93\u5165\u6570\u636e\u3001\u914d\u7f6e\u548c\u90e8\u7f72\u6587\u4ef6\u3002\u7f3a\u5931\u7684\u6d4b\u8bd5\u3001\u6e05\u5355\u6216\u5305\u521d\u59cb\u5316\u6587\u4ef6\u901a\u5e38\u8868\u793a\u6e90\u6811\u4e0d\u5b8c\u6574\u3002<\/p>\n<h3>C \u91cd\u65b0\u67e5\u8be2\u7ec6\u8282<\/h3>\n<p>\u5728\u6240\u6709\u56db\u79cd\u53d8\u4f53\u4e2d&#xff0c;\u751f\u6210\u7684\u8bf7\u6c42\u6307\u5b9a\u4e00\u4e2a\u53ef\u89c2\u5bdf\u7684\u76ee\u6807&#xff0c;\u540c\u65f6\u5c06\u5b9e\u73b0\u7b56\u7565\u7559\u7ed9\u7f16\u7801\u667a\u80fd\u4f53\u3002<\/p>\n<p>\u8de8\u5de5\u4f5c\u7a7a\u95f4\u5e03\u5c40\u3002\u00a0\u6bcf\u4e2a\u8de8\u5de5\u4f5c\u7a7a\u95f4\u5bf9\u4f5c\u4e3a repo0 \u548c repo1 \u6302\u8f7d\u5728 \/app\/workspaces \u4e0b\u3002\u751f\u6210\u667a\u80fd\u4f53\u9009\u62e9\u4e00\u4e2a\u53ef\u5199\u7684\u76ee\u6807\u548c\u53ea\u8bfb\u7684\u53c2\u8003\u3002\u7f16\u7801\u667a\u80fd\u4f53\u770b\u5230\u76f8\u540c\u7684\u5e03\u5c40&#xff0c;\u9a8c\u8bc1\u5668\u68c0\u67e5\u53c2\u8003\u4fdd\u6301\u4e0d\u53d8 (\u00a73.3)\u3002<\/p>\n<p>\u63d0\u793a\u5c55\u793a\u3002\u00a0\u6a21\u677f\u4fdd\u7559\u6bcf\u4e2a\u751f\u6210\u5668\u7684\u8f93\u5165\u3001\u6307\u4ee4\u548c\u8f93\u51fa\u6a21\u5f0f\u3002\u91cd\u590d\u7684\u89c4\u5219\u548c\u793a\u4f8b\u88ab\u7f29\u77ed\u3002\u5c16\u62ec\u53f7\u4e2d\u7684\u503c\u5728\u8fd0\u884c\u65f6\u586b\u5145\u3002<\/p>\n<h4>C.1 \u610f\u56fe\u6062\u590d<\/h4>\n<p>\u751f\u6210\u673a\u5236\u3002\u00a0\u610f\u56fe\u6062\u590d\u63a5\u6536\u6e05\u7406\u540e\u7684\u4efb\u52a1\u3001\u65f6\u95f4\u987a\u5e8f\u7684\u5bf9\u8bdd\u8f68\u8ff9\u548c\u6587\u4ef6\u8bc1\u636e\u3002\u5355\u8f6e\u8f68\u8ff9\u76f4\u63a5\u4f7f\u7528\u5176\u552f\u4e00\u7684\u5b9e\u8d28\u6027\u7528\u6237\u8bf7\u6c42\u3002\u5bf9\u4e8e\u591a\u8f6e\u8f68\u8ff9&#xff0c;\u7b2c\u4e00\u4e2a\u5b9e\u8d28\u6027\u7528\u6237\u8bf7\u6c42\u8bbe\u5b9a\u4efb\u52a1\u4e3b\u9898&#xff1b;\u540e\u7eed\u8bf7\u6c42\u4ec5\u5728\u5b83\u4eec\u6f84\u6e05\u3001\u7ea6\u675f\u6216\u6269\u5c55\u540c\u4e00\u4efb\u52a1\u65f6\u88ab\u7eb3\u5165&#xff0c;\u4e0d\u76f8\u5173\u7684\u4efb\u52a1\u8f6c\u79fb\u88ab\u6392\u9664\u3002\u6211\u4eec\u4f7f\u7528\u667a\u80fd\u4f53\u884c\u52a8\u548c\u6587\u4ef6\u8bc1\u636e\u6765\u89e3\u91ca\u8bf7\u6c42&#xff0c;\u4f46\u53ea\u4fdd\u7559\u7528\u6237\u9648\u8ff0\u7684\u9700\u6c42\u3002\u53ea\u6709\u751f\u6210\u7684 core_objective \u6210\u4e3a\u610f\u56fe\u6062\u590d\u6307\u4ee4&#xff1b;\u5176\u4f59\u8f93\u51fa\u5b57\u6bb5\u4f5c\u4e3a\u5143\u6570\u636e\u4fdd\u7559\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u4e00\u3001\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898<br \/>\n\u6838\u5fc3\u77db\u76fe \u8f68\u8ff9\u4e30\u5bcc&#xff0c;\u73af\u5883\u7a00\u7f3a&#xff1a;\u7ec8\u7aef\u4ee3\u7801\u667a\u80fd\u4f53\u4ea7\u751f\u5927\u91cf\u8f68\u8ff9\u6570\u636e&#xff0c;\u4f46\u771f\u5b9e\u3001\u53ef\u6267\u884c\u7684\u73af\u5883\u5374\u5f88\u5c11  \u8f68\u8ff9 vs \u73af\u5883&#xff1a;\u8f68\u8ff9\u662f\u56fa\u5b9a\u7684\u5355\u6b21\u6f14\u793a&#xff0c;\u8d28\u91cf\u53d7\u9650\u4e8e\u751f\u6210\u6a21\u578b&#xff1b;\u73af\u5883\u53ef\u91cd\u590d\u67e5\u8be2\u3001\u9a8c\u8bc1\u3001\u751f\u6210\u66f4\u96be\u4efb\u52a1&#xff0c;\u662f\u540e\u8bad\u7ec3\u771f\u6b63\u9700\u8981\u7684\u8d44\u6e90  \u4eba\u5de5\u6784\u5efa\u73af\u5883\u7684\u74f6\u9888&#xff1a;\u4e13\u5bb6\u6784\u5efa\u7684\u57fa\u51c6&#xff08;\u5982 Terminal-Bench&#xff09;\u8d28\u91cf\u9ad8\u4f46\u6570\u91cf\u6709\u9650&#xff0c;\u65e0<\/p>\n","protected":false},"author":2,"featured_media":101097,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[11557,11558,11556,11555,74],"topic":[],"class_list":["post-101106","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-server","tag-terminal","tag-universe","tag-11556","tag-11555","tag-agent"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v20.3 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments\u2014\u2014\u5c06\u667a\u80fd\u4f53\u8f68\u8ff9\u8f6c\u5316\u4e3a\u53ef\u6269\u5c55\u7684\u7ec8\u7aef\u73af\u5883 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3<\/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:\/\/www.wsisp.com\/helps\/101106.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments\u2014\u2014\u5c06\u667a\u80fd\u4f53\u8f68\u8ff9\u8f6c\u5316\u4e3a\u53ef\u6269\u5c55\u7684\u7ec8\u7aef\u73af\u5883 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"\u4e00\u3001\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898 \u6838\u5fc3\u77db\u76fe \u8f68\u8ff9\u4e30\u5bcc&#xff0c;\u73af\u5883\u7a00\u7f3a&#xff1a;\u7ec8\u7aef\u4ee3\u7801\u667a\u80fd\u4f53\u4ea7\u751f\u5927\u91cf\u8f68\u8ff9\u6570\u636e&#xff0c;\u4f46\u771f\u5b9e\u3001\u53ef\u6267\u884c\u7684\u73af\u5883\u5374\u5f88\u5c11 \u8f68\u8ff9 vs \u73af\u5883&#xff1a;\u8f68\u8ff9\u662f\u56fa\u5b9a\u7684\u5355\u6b21\u6f14\u793a&#xff0c;\u8d28\u91cf\u53d7\u9650\u4e8e\u751f\u6210\u6a21\u578b&#xff1b;\u73af\u5883\u53ef\u91cd\u590d\u67e5\u8be2\u3001\u9a8c\u8bc1\u3001\u751f\u6210\u66f4\u96be\u4efb\u52a1&#xff0c;\u662f\u540e\u8bad\u7ec3\u771f\u6b63\u9700\u8981\u7684\u8d44\u6e90 \u4eba\u5de5\u6784\u5efa\u73af\u5883\u7684\u74f6\u9888&#xff1a;\u4e13\u5bb6\u6784\u5efa\u7684\u57fa\u51c6&#xff08;\u5982 Terminal-Bench&#xff09;\u8d28\u91cf\u9ad8\u4f46\u6570\u91cf\u6709\u9650&#xff0c;\u65e0\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/101106.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-09-05T23:26:30+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260905232623-6a9ca51f9477e.png\" \/>\n<meta name=\"author\" content=\"admin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"9 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/101106.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/101106.html\",\"name\":\"Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments\u2014\u2014\u5c06\u667a\u80fd\u4f53\u8f68\u8ff9\u8f6c\u5316\u4e3a\u53ef\u6269\u5c55\u7684\u7ec8\u7aef\u73af\u5883 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-09-05T23:26:30+00:00\",\"dateModified\":\"2026-09-05T23:26:30+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/101106.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/101106.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/101106.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments\u2014\u2014\u5c06\u667a\u80fd\u4f53\u8f68\u8ff9\u8f6c\u5316\u4e3a\u53ef\u6269\u5c55\u7684\u7ec8\u7aef\u73af\u5883\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\",\"url\":\"https:\/\/www.wsisp.com\/helps\/\",\"name\":\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"description\":\"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\",\"name\":\"admin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"contentUrl\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"caption\":\"admin\"},\"sameAs\":[\"http:\/\/wp.wsisp.com\"],\"url\":\"https:\/\/www.wsisp.com\/helps\/author\/admin\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments\u2014\u2014\u5c06\u667a\u80fd\u4f53\u8f68\u8ff9\u8f6c\u5316\u4e3a\u53ef\u6269\u5c55\u7684\u7ec8\u7aef\u73af\u5883 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","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:\/\/www.wsisp.com\/helps\/101106.html","og_locale":"zh_CN","og_type":"article","og_title":"Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments\u2014\u2014\u5c06\u667a\u80fd\u4f53\u8f68\u8ff9\u8f6c\u5316\u4e3a\u53ef\u6269\u5c55\u7684\u7ec8\u7aef\u73af\u5883 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"\u4e00\u3001\u7814\u7a76\u80cc\u666f\u4e0e\u95ee\u9898 \u6838\u5fc3\u77db\u76fe \u8f68\u8ff9\u4e30\u5bcc&#xff0c;\u73af\u5883\u7a00\u7f3a&#xff1a;\u7ec8\u7aef\u4ee3\u7801\u667a\u80fd\u4f53\u4ea7\u751f\u5927\u91cf\u8f68\u8ff9\u6570\u636e&#xff0c;\u4f46\u771f\u5b9e\u3001\u53ef\u6267\u884c\u7684\u73af\u5883\u5374\u5f88\u5c11 \u8f68\u8ff9 vs \u73af\u5883&#xff1a;\u8f68\u8ff9\u662f\u56fa\u5b9a\u7684\u5355\u6b21\u6f14\u793a&#xff0c;\u8d28\u91cf\u53d7\u9650\u4e8e\u751f\u6210\u6a21\u578b&#xff1b;\u73af\u5883\u53ef\u91cd\u590d\u67e5\u8be2\u3001\u9a8c\u8bc1\u3001\u751f\u6210\u66f4\u96be\u4efb\u52a1&#xff0c;\u662f\u540e\u8bad\u7ec3\u771f\u6b63\u9700\u8981\u7684\u8d44\u6e90 \u4eba\u5de5\u6784\u5efa\u73af\u5883\u7684\u74f6\u9888&#xff1a;\u4e13\u5bb6\u6784\u5efa\u7684\u57fa\u51c6&#xff08;\u5982 Terminal-Bench&#xff09;\u8d28\u91cf\u9ad8\u4f46\u6570\u91cf\u6709\u9650&#xff0c;\u65e0","og_url":"https:\/\/www.wsisp.com\/helps\/101106.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-09-05T23:26:30+00:00","og_image":[{"url":"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260905232623-6a9ca51f9477e.png"}],"author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"9 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/101106.html","url":"https:\/\/www.wsisp.com\/helps\/101106.html","name":"Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments\u2014\u2014\u5c06\u667a\u80fd\u4f53\u8f68\u8ff9\u8f6c\u5316\u4e3a\u53ef\u6269\u5c55\u7684\u7ec8\u7aef\u73af\u5883 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-09-05T23:26:30+00:00","dateModified":"2026-09-05T23:26:30+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/101106.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/101106.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/101106.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments\u2014\u2014\u5c06\u667a\u80fd\u4f53\u8f68\u8ff9\u8f6c\u5316\u4e3a\u53ef\u6269\u5c55\u7684\u7ec8\u7aef\u73af\u5883"}]},{"@type":"WebSite","@id":"https:\/\/www.wsisp.com\/helps\/#website","url":"https:\/\/www.wsisp.com\/helps\/","name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","description":"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41","name":"admin","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/","url":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","contentUrl":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","caption":"admin"},"sameAs":["http:\/\/wp.wsisp.com"],"url":"https:\/\/www.wsisp.com\/helps\/author\/admin"}]}},"_links":{"self":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/101106","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/comments?post=101106"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/101106\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media\/101097"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=101106"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=101106"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=101106"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=101106"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}