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。

Teaching LLMs to Refuse Unknown Questions (TLDR: Train on "I Dunno" as Well)

2 点作者 wizzard0超过 1 年前

1 comment

wizzard0超过 1 年前
&gt; Our research is motivated by the observation that previous instruction tuning methods force the model to complete a sentence no matter whether the model knows the knowledge or not.<p>And so... adding the random samples that include &quot;I don&#x27;t know&quot; to the training datasets brings the accuracy on unknown questions from &lt;10% to over 90%