The Decoder· Jonathan Kemper·· 5 小时前AI 评分53
阿里与亚洲多所高校发布Co-Design Bench编码智能体评测基准
AI coding agents still need human cleanup on nearly three out of four tasks in a real-world test
AI 导读
阿里与亚洲多所高校研究人员发布Co-Design Bench,一个基于真实用户变更请求的编码智能体评测基准。测试显示,即使表现最好的GPT-6 Astra,仅在26.4%的任务中完全正确完成,平均满足70%需求。
来源:The Decoder · the-decoder.com