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。

Learning to summarize from human feedback (2022)

56 点作者 georgehill超过 2 年前

5 条评论

nl超过 2 年前
Note that this is pretty old (2020).<p>They released code, models and raw data here: <a href="https:&#x2F;&#x2F;github.com&#x2F;openai&#x2F;summarize-from-feedback">https:&#x2F;&#x2F;github.com&#x2F;openai&#x2F;summarize-from-feedback</a>
O__________O超过 2 年前
Feel like automating the human feedback, not the summaries themselves, that should have been the core focus of research like this. As is, even reading guidelines for summary evaluation they presented the reviewers are not reproducible.
Andugal超过 2 年前
Sorry if the answer is obvious but can we use this for our own usage? If yes, how?
basch超过 2 年前
I&#x27;d be curious to see how this does compared to models trained on more professional datasets than reddit tldr.<p>For example, train a model(s) by reading every single article (including paywall&#x2F;cache replacement) of <a href="https:&#x2F;&#x2F;www.techmeme.com&#x2F;river" rel="nofollow">https:&#x2F;&#x2F;www.techmeme.com&#x2F;river</a> <a href="https:&#x2F;&#x2F;www.mediagazer.com&#x2F;river" rel="nofollow">https:&#x2F;&#x2F;www.mediagazer.com&#x2F;river</a> <a href="https:&#x2F;&#x2F;www.memeorandum.com&#x2F;river" rel="nofollow">https:&#x2F;&#x2F;www.memeorandum.com&#x2F;river</a> <a href="https:&#x2F;&#x2F;www.wesmirch.com&#x2F;river" rel="nofollow">https:&#x2F;&#x2F;www.wesmirch.com&#x2F;river</a> <a href="https:&#x2F;&#x2F;ballbug.com&#x2F;river" rel="nofollow">https:&#x2F;&#x2F;ballbug.com&#x2F;river</a> and comparing it to the summary headline.
评论 #35027719 未加载
评论 #35026090 未加载
marckohlbrugge超过 2 年前
Does anyone have a TL;DR on this?
评论 #35027123 未加载