TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Perfect usecase for a bad LLM models

2 点作者 shsethi3 个月前

1 comment

genewitch3 个月前
I think there will be some argument in the comments about whether any LLM can be considered unusable for this test.<p>That is, is there any current LLM where you could prompt it with something like an interview question and it would give you perfect code one shot or zero shot?<p>It might be better to host an LLM and craft the questions such that you know that the LLM will screw it up. Or, since you&#x27;re self-hosting, you can have the system prompt, say insert subtle bugs that aren&#x27;t syntax errors. I&#x27;ll have to test if that actually does anything.
评论 #43301527 未加载