前两天换成了 Copilot Pro+，然后在 VS Code 里用 Cline 插件接入 Copilot Pro+，今天看到有人在推上说 Claude 和 GPT 都已经降智到无法通过洗车测试，然后就试着测了一遍，结果从 Opus 4.6 到 Sonnet 4.6 再到 GPT 5.4 ，全部都无法通过洗车测试，提问：I want to wash my car. The car wash is 50m away, should I walk or drive?全部都回答应该走过去。
用另一家模型聚合平台在网页上测试，Opus 4.6 又测试通过并且连测 5 次结果都稳定（不过 Sonnet 4.6 和 GPT 5.4 还是失败），这是咋回事。
 | 1 Maybez007 1 小时 30 分钟前 I want to wash my car. The car wash is 50m away, should I walk or drive? Walk. 50 m is roughly a one-minute walk, and driving that distance is more hassle than it’s worth. Exceptions: drive only if the weather is bad, you have mobility issues, or you need to bring bulky items. are you sure? Yes. For 50 meters, walking is the practical choice. Driving that distance usually means: - getting in and out of the car - starting it and maneuvering - possibly waiting or repositioning anyway That often takes longer than just walking there. If you mean you need to bring the car to the wash, then drive the car to the wash and walk home only if they keep it for a while. OpenAI 官网订阅的 gpt-5.4 的回复...... |
Select Language