**真实测试显示:AI 编程代理在四分之三任务中仍需人工清理** _AI coding agents still need human cleanup on
原帖
**真实测试显示:AI 编程代理在四分之三任务中仍需人工清理**
_AI coding agents still need human cleanup on nearly three out of four tasks in a real-world test_
> 阿里巴巴发布的新型基准测试 Co-Design Bench 针对真实用户请求,结果显示即使是表现最佳的 GPT-6 Astra 模型,也仅能完全解决 26.4% 的任务。这表明在现实世界应用中,AI 编程代理生成的代码在绝大多数情况下仍需要人工进行清理和修正。
**来源信息**
- **来源**:The Decoder:AI News(RSS)
- **分类**:ai-models
- **发布时间**:2026-11-08 17:50(北京时间)
- **原文**:[打开原文](https://the-decoder.com/ai-coding-agents-still-need-human-cleanup-on-nearly-three-out-of-four-tasks-in-a-real-world-test)