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。

Show HN: The Calibration game – get better at identifying hallucination in LLMs

2 点作者 viktour19超过 1 年前
Calibration game is a game to get better at identifying hallucination in LLMs.<p>Prompts and hallucination labels (using ChatGPT) are obtained from Alpaca and HaluEval. You can use your own dataset to calibrate users to the responses of a different LLM (<a href="https:&#x2F;&#x2F;github.com&#x2F;Viktour19&#x2F;calibrationgame">https:&#x2F;&#x2F;github.com&#x2F;Viktour19&#x2F;calibrationgame</a>).<p>ML engineers are starting to come around to the fact that hallucination in LLMs is really tough to mitigate. This is an active area of research and we may have LLMs with no hallucination in the future. At the moment, we need to design user experiences that align users with this behaviour. The hope with the Calibration game is that it equips engineers to calibrate users to the responses of LLMs used in their applications, e.g., as part of the onboarding process.<p>Would love to hear some feedback.

1 comment

tamimio超过 1 年前
Sounds like free labor to fix someone’s LLM