{"id":12133,"date":"2024-06-25T17:54:49","date_gmt":"2024-06-25T09:54:49","guid":{"rendered":"https:\/\/ihower.tw\/blog\/?p=12133"},"modified":"2025-07-19T19:29:25","modified_gmt":"2025-07-19T11:29:25","slug":"aie-new-software-era","status":"publish","type":"post","link":"https:\/\/ihower.tw\/blog\/12133-aie-new-software-era","title":{"rendered":"\u611b\u597d AI Engineer \u96fb\u5b50\u5831 \ud83d\ude80\u00a0\u65b0\u7684\u8edf\u9ad4\u958b\u767c\u6642\u4ee3\u4f86\u81e8 #13"},"content":{"rendered":"\n
\"\"<\/figure>\n\n\n\n

Hello! \u4f60\u597d \ud83d\udc4b<\/p>\n\n\n\n

\u6211\u662f ihower\uff0c\u9019\u671f\u958b\u59cb\u6709\u500b\u65b0\u7cfb\u5217\u662f Google Colab \u5206\u4eab\uff0c\u6211\u6703\u88fd\u4f5c\u65e2\u5be6\u7528\u53c8\u5e36\u6709\u6559\u5b78\u610f\u7fa9\u7684\u5c0f\u7a0b\u5f0f\u7d66\u5927\u5bb6\u3002<\/p>\n\n\n\n

\ud83d\udd1dWith AI, Anyone Can Be a Coder Now | Thomas Dohmke | TED<\/a><\/h3>\n\n\n\n

\u8981\u5982\u4f55\u7406\u89e3\u751f\u6210\u5f0f AI \u6703\u5982\u4f55\u5f71\u97ff\u8edf\u9ad4\u958b\u767c?<\/p>\n\n\n\n\n\n\n\n

GitHub \u662f\u76ee\u524d\u505a AI coding \u6700\u53b2\u5bb3\u7684\u516c\u53f8\uff0c\u4e0d\u4f46\u6709\u5168\u4e16\u754c\u6700\u591a\u7684 source code \u53ef\u4ee5\u8a13\u7df4\uff0c\u4e5f\u6709\u5fae\u8edf\u7238\u7238\u7684\u7b97\u529b\u3002
\u6700\u8fd1\u4ed6\u5011 CEO Thomas Dohmke \u6709\u5834 TED \u79d1\u666e\u6f14\u8b1b: \u900f\u904e AI \u4efb\u4f55\u4eba\u90fd\u53ef\u4ee5\u6210\u70ba Coder (15\u5206\u9418\u3001\u6709\u7c21\u4e2d\u5b57\u5e55) \u503c\u5f97\u4e00\u770b\u3002<\/p>\n\n\n\n

– \u76ee\u524d\u53ef\u7528 Github Copilot \u5de5\u5177\u5354\u52a9\u4f60\u5b8c\u6210\u7a0b\u5f0f\u78bc\uff0c\u5c07\u9020\u798f\u66f4\u591a\u4eba\u53ef\u4ee5\u8f15\u9b06\u958b\u59cb\u5beb\u7a0b\u5f0f
– \u900f\u904e\u81ea\u7136\u8a9e\u8a00\u8207\u7a0b\u5f0f\u78bc\u4e92\u52d5\uff0c\u672a\u4f86\u5c07\u6709\u66f4\u591a\u4eba\u53ef\u4ee5\u662f Coder
– \u76ee\u524d Github \u4e0a\u6709 100M \u958b\u767c\u8005\uff0c\u9810\u8a08\u5230 2030 \u5e74\u6210\u9577\u5341\u500d\uff0c\u4e5f\u5c31\u662f 1B \u7684\u958b\u767c\u8005\u3002\u5168\u7403\u4eba\u53e3 10% \u4e0d\u53ea\u6703\u7528\u96fb\u8166\uff0c\u9084\u6703\u5beb\u7a0b\u5f0f!
– \u6700\u5f8c\u9084 demo \u4e86\u6700\u65b0\u9810\u89bd\u7248\u529f\u80fd Github Workspace<\/a>: \u53ea\u8981\u958b\u7968\uff0cGithub \u5c31\u53ef\u4ee5\u5e6b\u4f60\u62c6\u89e3\u6210\u6587\u5b57\u898f\u683c\uff0c\u7d93\u904e\u4f60\u78ba\u8a8d\u4fee\u6539\u5f8c\uff0c\u518d\u5e6b\u4f60\u5beb code \u7522\u751f pull request\uff0c\u6700\u5f8c\u4f60\u518d\u6aa2\u67e5\u7a0b\u5f0f\u78bc\u5408\u4f75\u9032\u4e3b\u7a0b\u5f0f
– \u6700\u5f8c\u4e3b\u6301\u4eba\u63d0\u554f\u4e00\u500b\u95dc\u9375\u554f\u984c:   \u9810\u6e2c\u672a\u4f86\u662f\u5426\u9084\u9700\u8981\u4eba\u985e\u53c3\u8207? \u56de\u7b54\u662f: \u5c08\u696d\u8edf\u9ad4\u8a2d\u8a08\u5e2b\u4e0d\u6703\u6d88\u5931\uff0c\u4ecd\u9700\u8981\u4eba\u985e\u53bb\u7406\u89e3\u8981\u6253\u9020\u4ec0\u9ebc\u6771\u897f\uff0c\u4ecd\u9700\u8981\u4eba\u985e\u5c08\u696d\u53bb\u7dad\u8b77\u8f03\u5927\u7684\u8907\u96dc\u8edf\u9ad4\u7cfb\u7d71\u3002<\/p>\n\n\n\n

\u5e7e\u9ede\u81ea\u5df1\u7684\u770b\u6cd5:<\/p>\n\n\n\n

1. \u6574\u5834 talk \u53ea\u8b1b Copilot (AI\u5354\u52a9\u4eba\u985e\u5bebcode)\uff0c\u6c92\u6709\u63d0\u5230 Autopilot (\u5168\u81ea\u52d5\u958b\u767c\u8edf\u9ad4)\uff0c\u6211\u89ba\u5f97\u9019\u624d\u662f\u73fe\u6cc1\u5be6\u969b\u53ef\u884c\u7684\u767c\u5c55\u3002\u8b1b\u4ec0\u9ebc AI \u6703\u53d6\u4ee3\u8edf\u9ad4\u5de5\u7a0b\u5e2b\uff0c\u5c08\u696d\u4e0d\u8981\u5b78\u8cc7\u8a0a\u5de5\u7a0b\u4e86\uff0c\u9019\u90fd\u53ea\u662f\u5f15\u767c FOMO \u7126\u616e\u6c92\u5565\u5e6b\u52a9\uff0c\u5c31\u8ddf\u53bb\u731c\u660e\u5e74\u5f8c\u5e74\u5c31\u6709 AGI \u4e00\u6a23\u3002<\/p>\n\n\n\n

2. \u6211\u8a8d\u70ba\u751f\u6210\u5f0f AI \u4e0d\u4f46\u4e0d\u6703\u6e1b\u5c11\u8edf\u9ad4\u5de5\u7a0b\u5e2b\u7684\u9700\u6c42\uff0c\u53cd\u800c\u6703\u589e\u52a0\u3002\u6709\u500b\u73fe\u8c61\u53eb\u505a “Jevons paradox” (\u6770\u6587\u65af\u6096\u8ad6<\/a>): \u7576\u67d0\u9805\u8cc7\u6e90\u7684\u4f7f\u7528\u6548\u7387\u63d0\u9ad8\u6642\uff0c\u53cd\u800c\u53ef\u80fd\u5c0e\u81f4\u5c0d\u8a72\u8cc7\u6e90\u9700\u6c42\u7684\u589e\u52a0\uff0c\u800c\u4e0d\u662f\u6e1b\u5c11\u3002\u7576\u4eba\u4eba\u90fd\u53ef\u4ee5\u5beb\u9ede\u7a0b\u5f0f\u7684\u6642\u5019\uff0c\u5404\u884c\u5404\u696d\u5c31\u6703\u958b\u59cb\u81ea\u884c\u505a\u4e9b\u81ea\u52d5\u5316\u7a0b\u5f0f\uff0c\u9019\u5c07\u6703\u7522\u751f\u66f4\u591a IT \u8edf\u9ad4\u9700\u6c42\u3002<\/p>\n\n\n\n

\u4f8b\u5982\u6700\u8fd1 Muyueh Lee \u5c31\u5728\u5e6b\u52a9\u4e00\u822c\u4e0a\u73ed\u65cf\u300c\u7528 AI \u5beb\u7a0b\u5f0f\uff0c\u8b93\u81ea\u5df1\u65e9\u4e0b\u73ed<\/a>\u300d\uff0cAI \u5c0f\u805a\u8ddf\u5e74\u6703\u4e5f\u5e38\u6709\u516c\u53f8\u5206\u4eab\u4ed6\u5011\u5982\u4f55\u81ea\u5df1\u505a\u81ea\u52d5\u5316\u3002\u9019\u6a23\u6f14\u9032\u4e0b\u53bb\uff0c\u7576\u55ae\u5175\u4e0d\u6eff\u8db3\u65bc “\u81ea\u52d5\u5316\u7a0b\u5f0f”\uff0c\u60f3\u8981\u505a\u4e00\u500b\u9577\u671f\u7dad\u904b\u7684\u8907\u96dc “\u8edf\u9ad4\u7cfb\u7d71” \u6642\uff0c\u5c31\u6703\u9700\u8981\u5c08\u696d\u8edf\u9ad4\u5de5\u7a0b\u5e2b\u7684\u5354\u52a9\u3002<\/p>\n\n\n\n

3. \u5c0d\u8edf\u9ad4\u5de5\u7a0b\u5e2b\u4f86\u8aaa\uff0c\u6280\u80fd\u7d44\u80af\u5b9a\u662f\u6703\u6539\u8b8a\u7684\u3002\u9019\u4ef6\u4e8b\u60c5\u6211\u60f3\u5927\u5bb6\u90fd\u5df2\u7d93\u662f\u904e\u4f86\u4eba\uff0c\u820a\u6280\u80fd\u7684\u9700\u6c42\u6703\u6e1b\u5c11\uff0c\u65b0\u6280\u80fd\u7684\u9700\u6c42\u6703\u589e\u52a0\u3002\u9019\u5e7e\u5929\u53c8\u4e0a\u4e86\u4e00\u6b21 Will \u4fdd\u54e5 \u7684 Github Copilot \u8ab2\u7a0b\uff0c\u4fdd\u54e5\u5c31\u8b1b\u4ed6\u73fe\u5728\u7528\u5634\u5df4\u5beb\u7a0b\u5f0f\uff0c\u6253\u5b57\u9000\u5316\uff0c\u56e0\u70ba Github Copilot \u53ef\u4ee5\u63a5\u53d7\u8a9e\u97f3\u8fa8\u8b58\u8f38\u5165\u6307\u4ee4\u5354\u52a9\u4f60\u5beb code\u3002\u9019\u6a23\u900f\u904e\u8207 AI Agent \u4ea4\u4e92\u5354\u4f5c\u958b\u767c\u8edf\u9ad4\u7684 UX \u9ad4\u9a57\uff0c\u662f\u76ee\u524d\u6f14\u9032\u4e2d\u7684\u8da8\u52e2\u3002<\/p>\n\n\n\n

\u6240\u4ee5\uff0c\u53ea\u8981\u4f60\u5c0d coding \u6709\u8208\u8da3\uff0c\u73fe\u5728\u5b78\u662f\u518d\u597d\u4e0d\u904e\u7684\u6642\u6a5f\u3002\u4eab\u53d7\u65b0\u7684\u8edf\u9ad4\u958b\u767c\u6642\u4ee3\u4f86\u81e8\u5427\u3002<\/p>\n\n\n\n

\u5c0d Autopilot Coding \u767c\u5c55\u6709\u8208\u8da3\u7684\uff0c\u53ef\u4ee5\u95dc\u6ce8 SWE-bench \u8a55\u6e2c<\/a>(AI \u662f\u5426\u80fd\u5168\u81ea\u52d5\u89e3\u6c7a github issue) \u76ee\u524d\u6700\u597d\u7684\u53ea\u6709 19.27 \u6210\u529f\u7387\u3002\u6240\u4ee5\u4f60\u8981\u8aaa AI \u5df2\u7d93\u666e\u904d\u8d85\u8d8a\u521d\u968e\u8edf\u9ad4\u5de5\u7a0b\u5e2b\u6211\u662f\u4e0d\u76f8\u4fe1\u5566\u3002\u9019\u6709\u4eba\u505a\u6df1\u5165\u5206\u6790 How Do AI Software Engineers Really Compare To Humans?<\/a>\uff0c\u4f60\u6703\u767c\u73fe\u80fd\u89e3\u6c7a\u7684 issue \u5927\u90e8\u5206\u90fd\u662f\u5c0f\u7bc4\u570d\u7684\u4fee\u6539\u3002<\/p>\n\n\n\n

\ud83d\udc4dApple Intelligence \u767c\u8868<\/a><\/h3>\n\n\n\n

\u4eca\u5e74 WWDC \u773e\u6240\u671f\u5f85\u767c\u8868\u4e86 Apple Intelligence\uff0c\u770b\u4f86\u4e3b\u8981\u662f\u7531\u4e09\u500b LLM \u6240\u7d44\u6210:<\/p>\n\n\n\n

– Apple \u5730\u7aef Siri: \u5224\u65b7\u554f\u984c\u985e\u578b\u3001\u57fa\u672c\u6587\u5b57\u4efb\u52d9\u3001\u547c\u53eb\u672c\u6a5f function \u7b49\u7c21\u55ae\u4efb\u52d9\uff0c\u9019\u662f\u500b Apple \u81ea\u88fd\u7684 3B \u5c0f\u6a21\u578b\u3002\u53e6\u5916\u672c\u6a5f\u9084\u8981\u8ca0\u8cac\u505a semantic search \u627e\u51fa\u8981\u53c3\u8003\u7684\u500b\u4eba context \u8cc7\u6599\u3002
– Apple \u96f2\u7aef Private Cloud Compute \u7684 LLM: \u7576\u7528\u6236\u6307\u4ee4\u6bd4\u8f03\u8907\u96dc\u6642\u6539\u7528\u9019\u500b\uff0c\u6703\u6839\u64da\u50b3\u4e0a\u4f86\u7684 context \u8cc7\u6599\u505a\u8907\u96dc\u4efb\u52d9\u898f\u5283\uff0c\u9019\u662f\u500b Apple \u81ea\u88fd\u7279\u8abf\u6b64\u5834\u666f\u7684\u5927\u6a21\u578b (\u9019 server \u4e5f\u662f\u7528 Apple Silicon \u6676\u7247\uff0c\u4e0d\u662f Nvidia \u5594\uff09
– \u6574\u5408\u5916\u90e8 OpenAI ChatGPT: \u6703\u7528\u65bc\u6cdb\u7528\u7684\u77e5\u8b58\u554f\u7b54\u804a\u5929\uff0c\u7576 Siri \u5224\u65b7\u7528\u6236\u7684\u554f\u984c\u9069\u5408 gpt-4o \u6642\uff0c\u6703\u8df3\u63d0\u9192\u554f\u7528\u6236\u662f\u5426\u5141\u8a31\u53bb\u554f OpenAI\u3002\u9019\u500b\u5916\u90e8\u6a21\u578b\u6211\u60f3\u4e0d\u540c\u5730\u5340\u5c07\u6703\u662f\u53ef\u5207\u63db\u7684\u3002<\/p>\n\n\n\n

\u8a73\u7d30\u6280\u8853\u53ef\u4ee5\u770b\u5b98\u65b9\u8cc7\u6599 Introducing Apple\u2019s On-Device and Server Foundation Models<\/a> \u548c Platforms State of the Union \u5f71\u7247<\/a>(0:01:35~0:09:14 \u9019\u6bb5\u6709\u8aaa\u660e\u67b6\u69cb)
\u6700\u91cd\u8981\u7684\u5c31\u662f\u9019\u5f35\u67b6\u69cb\u5716\u4e86:<\/p>\n\n\n\n

\"\"<\/a><\/figure>\n\n\n\n

\ud83c\udfaf\u674e\u5b8f\u6bc5\u6559\u6388 \u751f\u6210\u5f0f\u4eba\u5de5\u667a\u6167\u5c0e\u8ad6<\/a><\/h3>\n\n\n\n

\u53f0\u5927\u674e\u5b8f\u6bc5\u6559\u6388\u7684 “\u751f\u6210\u5f0f\u4eba\u5de5\u667a\u6167\u5c0e\u8ad6” \u8ab2\u7a0b\u5168\u90e8\u5b8c\u7d50\u5566! \u5305\u62ec\u8ab2\u7a0b\u5f71\u7247\u3001\u6295\u5f71\u7247\u548c\u4f5c\u696d \ud83e\udd29<\/p>\n\n\n\n

> \u9019\u9580\u8ab2\u7a0b\u65e8\u5728\u8b93\u5df2\u7d93\u719f\u6089\u751f\u6210\u5f0f AI \u5de5\u5177\uff08\u4f8b\u5982\uff1a\u7db2\u9801\u7248 ChatGPT\uff09\u7684\u540c\u5b78\uff0c\u6df1\u5165\u63a2\u7d22\u66f4\u591a\u751f\u6210\u5f0f AI \u7684\u6f5b\u529b\uff0c\u4e26\u638c\u63e1\u5176\u80cc\u5f8c\u7684\u539f\u7406\u3002\u9019\u9580\u8ab2\u5c0d\u5b78\u751f\u7684\u80cc\u666f\u6c92\u6709\u9810\u8a2d\uff0c\u4efb\u4f55\u80cc\u666f\u7684\u5b78\u751f\u90fd\u80fd\u7406\u89e3\u8ab2\u7a0b\u5167\u5bb9 \u3002<\/p>\n\n\n\n

\ud83d\udea7Google Colab \u5206\u4eab\u7cfb\u5217: LiteLLM \u5e73\u884c\u6e2c\u8a66\u591a\u5bb6\u6a21\u578b<\/a><\/h3>\n\n\n\n

\u5404\u4f4d LLM \u9ad8\u7d1a\u73a9\u5bb6\u5011\uff0c\u4e00\u5b9a\u5e38\u5e38\u60f3\u8981\u4e00\u6b21\u6e2c\u8a66\u591a\u5bb6\u6a21\u578b\u5427\u3002\u9019\u88cf\u5e6b\u5927\u5bb6\u628a Google Colab \u7a0b\u5f0f\u5beb\u597d\u4e86 (\u9700\u81ea\u5099 API Key)\u3002
\u4e00\u6b21\u4e32 28 \u500b LLM \u800c\u4e14\u662f\u5e73\u884c\u767c\u9001\uff0c\u901f\u5ea6\u5feb\u7684\u5148\u8fd4\u56de\u7d50\u679c\uff0c\u7e3d\u57f7\u884c\u6642\u9593 = \u6700\u6162\u7684\u90a3\u4e00\u500b<\/p>\n\n\n\n

\u7b2c\u4e00\u6b21\u8a2d\u5b9a\u6bd4\u8f03\u9ebb\u7169\u4e9b\uff0c\u4f60\u9700\u8981\u62ff\u5230\u5404\u5bb6\u7684 API Key \u8a2d\u5b9a\u4e00\u4e0b\uff0c  \u5305\u62ec OpenAI, Groq, Gemini, Anthropic, Cohere, Mistral, OpenRouter \u5171\u4e03\u5bb6  
\u90fd\u662f\u8a3b\u518a\u767b\u5165\u5f8c\u5c31\u53ef\u4ee5\u62ff\u5230\u4e86\uff0c\u9664\u4e86 OpenAI \u8981\u5148\u7d81\u4fe1\u7528\u5361\u624d\u6709 gpt-4\uff0c\u5176\u4ed6\u6211\u8a18\u5f97\u90fd\u4e0d\u9700\u8981\u3002\u5acc\u9ebb\u7169\u4f60\u5c31\u8a3b\u89e3\u6389\u4e0d\u9700\u8981\u7684\u6a21\u578b\u5373\u53ef\u3002  <\/p>\n\n\n\n

\u9806\u4fbf\u4ecb\u7d39\u4e00\u4e0b OpenRouter<\/a> \u9019\u5bb6 API Provider  
\u9019\u5bb6\u5176\u5be6\u662f\u500b API gateway\uff0c\u5f8c\u9762\u53c8\u4e32\u5f88\u591a\u5bb6\uff0c\u6709\u5e7e\u500b\u4e2d\u570b\u6a21\u578b\u9019\u88e1\u4e5f\u6709\uff0c\u4f60\u53ea\u9700\u8981\u7528\u9019\u500b API \u7aef\u9ede\u5c31\u53ef\u4ee5\u5c0d\u63a5\u591a\u5bb6\u6a21\u578b\u3002  
\u7136\u5f8c\u6211\u518d\u642d\u914d LiteLLM<\/a> \u9019\u500b Python \u5957\u4ef6\uff0c\u7d71\u4e00\u7528 OpenAI API \u683c\u5f0f\u5c31\u53ef\u4ee5\u5c0d\u63a5\u591a\u5bb6\u6a21\u578b\u3002  
\u6240\u4ee5 Litellm x OpenRouter \u5c31\u53ef\u4ee5\u8f15\u9b06\u5168\u90e8\u90fd\u6e2c\u5230\u5566\u3002<\/p>\n\n\n\n

\ud83d\udc4a\u4e0d\u516c\u958b\u6e2c\u8cc7\u7684 SEAL Leaderboards \u8a55\u6e2c<\/a><\/h3>\n\n\n\n

\u6709\u4e9b LLM \u7684 Benchmark \u8868\u73fe\u597d\uff0c\u662f\u56e0\u70ba\u5148\u770b\u4e86\u984c\u76ee\u628a\u7b54\u6848\u80cc\u8d77\u4f86\u4e86\uff0c\u800c\u4e0d\u662f\u8166\u5b50\u597d\u3002
\u4e0a\u500b\u6708 Scale AI \u7684\u4e00\u7bc7\u8ad6\u6587 Careful Examination of Large Language Model Performance on Grade School Arithmetic<\/a> \u5be6\u9a57\u8868\u660e\u4e86\u8cc7\u6599\u96c6\u6c61\u67d3\u537b\u6709\u5176\u4e8b\u3002<\/p>\n\n\n\n

\u6700\u5927\u82e6\u4e3b\u662f Mixtral 8x22B \u548c Phi-3\uff0c\u6e96\u78ba\u7387\u7684\u771f\u5be6\u8868\u73fe\u6bd4 Benchmark \u9084\u4f4e\u4e0a 13%\uff0c\u660e\u986f\u904e\u5ea6\u64ec\u5408(over-fitting) \u6709\u80cc\u66f8\u73fe\u8c61\u3002\u800c gpt-4 \u548c claude 3 \u8868\u73fe\u6700\u597d\uff0c\u771f\u5be6\u6027\u80fd\u66f4\u9ad8\u3002<\/p>\n\n\n\n

\u5be6\u9a57\u7684 Benchmark \u662f\u7528 GSM8k\uff0c\u9019\u662f\u4e00\u500b\u516c\u958b\u7684\u5c0f\u5b78\u7b97\u8853\u7684\u8cc7\u6599\u96c6\uff0c\u53ef\u4ee5\u60f3\u50cf\u5f88\u591a\u6a21\u578b\u5728\u8a13\u7df4\u7684\u6642\u5019\uff0c\u4e00\u5b9a\u6703\u628a\u9019\u500b\u984c\u76ee\u653e\u9032\u53bb\u8a13\u7df4\u3002\u800c Scale AI \u5f9e\u982d\u505a\u4e86\u4e00\u500b\u985e\u4f3c\u7684\u4e0d\u516c\u958b\u8cc7\u6599\u96c6 GSM1k \u4f86\u505a\u8a55\u6e2c\uff0c\u767c\u73fe\u5f88\u591a\u6a21\u578b\u7684 GSM1k \u5206\u6578\u6bd4 GSM8k \u9084\u4f4e! \u7279\u5225\u662f\u672c\u4f86\u5206\u6578\u5c31\u6bd4\u8f03\u4f4e(40%~70%)\u7684\u5c0f\u6a21\u578b\uff0c\u5176\u5be6\u969b\u7684\u6578\u5b78\u63a8\u7406\u80fd\u529b\u662f\u66f4\u4f4e\u7684\u3002<\/p>\n\n\n\n

Jim Fan \u5927\u795e\u5c31\u8aaa\u4e86<\/a>: \u9019\u7a2e\u4e0d\u516c\u958b\u8cc7\u6599\u96c6\u7684 benchmark\uff0c\u4f8b\u5982 Scale AI \u9019\u7a2e\u53ef\u9760\u7b2c\u4e09\u65b9\u505a\u7684\uff0c\u6703\u66f4\u53ef\u9760\u6709\u7528\uff0c\u56e0\u70ba\u6a21\u578b\u5ee0\u5546\u4e0d\u77e5\u9053\u984c\u76ee \ud83d\ude04<\/p>\n\n\n\n

\u800c\u9019\u500b SEAL Leaderboards<\/a> \u5c31\u662f Scale AI \u6700\u65b0\u63a8\u51fa\u7684\u8a55\u6e2c\u6392\u884c\u699c\uff0c\u984c\u76ee\u662f\u4e0d\u516c\u958b\u7684\u3002<\/p>\n\n\n\n

\ud83d\udc4a\u6bcf\u6708\u66f4\u65b0\u6e2c\u8cc7\u7684 LiveBench \u8a55\u6e2c<\/a><\/h3>\n\n\n\n

\u70ba\u4e86\u907f\u514d\u5404\u5bb6\u6a21\u578b\u56e0\u70ba\u80cc\u66f8\u9020\u6210\u9ad8\u5206\u5047\u8c61\uff0c\u9019\u662f Abacus.AI \u548c Yann LeCunn \u5408\u4f5c\u63a8\u51fa\u7684\u65b0 LiveBench \u8a55\u6e2c\u3002
\u4e4b\u6240\u4ee5\u53eb\u505a Live \u5c31\u662f\u56e0\u70ba\u6bcf\u6708\u90fd\u6703\u66f4\u65b0\u554f\u984c\u96c6(\u4f86\u81ea\u6700\u65b0\u7684\u8cc7\u6599\uff0c\u4f8b\u5982\u65b0\u805e\u3001\u8ad6\u6587\u3001\u96fb\u5f71\u7b49)\uff0c\u56e0\u6b64\u5404\u5bb6\u6a21\u578b\u5728\u8a13\u7df4\u6642\u662f\u7121\u6cd5\u5148\u80cc\u8d77\u4f86\u7684\u3002<\/p>\n\n\n\n

LiveBench \u76ee\u524d\u5305\u542b 6 \u500b\u985e\u5225\u4e2d\u7684 18 \u500b\u591a\u6a23\u5316\u4efb\u52d9\uff0c\u4e26\u5c07\u96a8\u8457\u6642\u9593\u7684\u63a8\u79fb\u767c\u5e03\u66f4\u65b0\u66f4\u96e3\u7684\u4efb\u52d9\u3002
\u9019\u500b\u6708\u525b\u63a8\u51fa\u7684 Claude 3.5 Sonnet<\/a> \u76ee\u524d\u6392\u7b2c\u4e00\uff0c\u4f46\u4e5f\u53ea\u662f\u52c9\u5f37\u53ca\u683c 62 \u5206\u3002
\u66f4\u8b93\u6211\u611f\u5230\u9707\u9a5a\u7684\u662f\uff0c\u5c0f\u6a21\u578b SLM \u7684\u5206\u6578\u771f\u7684\u6158\u70c8\uff0c\u5206\u6578\u5f88\u4f4e\uff0c\u9084\u6709 0 \u5206\u7684\u3002<\/p>\n\n\n\n

\u9019\u7a2e<del>\u6b98\u9177<\/del>\u9ad8\u9451\u5225\u529b\u7684\u8a55\u6e2c\u6700\u559c\u6b61\u4e86\u3002<\/p>\n\n\n\n

\u2013 ihower<\/p>\n","protected":false},"excerpt":{"rendered":"

Hello! \u4f60\u597d \ud83d\udc4b \u6211\u662f ihower\uff0c\u9019\u671f\u958b\u59cb\u6709\u500b\u65b0\u7cfb\u5217\u662f Google Colab \u5206\u4eab\uff0c\u6211 … <\/p>\n

\u95b1\u8b80\u5168\u6587\u3008\u611b\u597d AI Engineer \u96fb\u5b50\u5831 \ud83d\ude80\u00a0\u65b0\u7684\u8edf\u9ad4\u958b\u767c\u6642\u4ee3\u4f86\u81e8 #13\u3009<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"jetpack_post_was_ever_published":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2}},"categories":[91],"tags":[],"class_list":["post-12133","post","type-post","status-publish","format-standard","hentry","category-aie","entry"],"jetpack_publicize_connections":[],"jetpack_featured_media_url":"","jetpack_shortlink":"https:\/\/wp.me\/p1q6tG-39H","jetpack_sharing_enabled":true,"jetpack_likes_enabled":true,"_links":{"self":[{"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/posts\/12133","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/comments?post=12133"}],"version-history":[{"count":5,"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/posts\/12133\/revisions"}],"predecessor-version":[{"id":13020,"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/posts\/12133\/revisions\/13020"}],"wp:attachment":[{"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/media?parent=12133"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/categories?post=12133"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/tags?post=12133"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}