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。

Serious issues in Llama 4 training. VP of AI at Meta resigned

7 点作者 devops000大约 1 个月前

3 条评论

chvid大约 1 个月前
<a href="https:&#x2F;&#x2F;x.com&#x2F;Ahmad_Al_Dahle&#x2F;status&#x2F;1909302532306092107" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;Ahmad_Al_Dahle&#x2F;status&#x2F;1909302532306092107</a><p>“ We&#x27;ve also heard claims that we trained on test sets -- that&#x27;s simply not true and we would never do that. Our best understanding is that the variable quality people are seeing is due to needing to stabilize implementations.”
chvid大约 1 个月前
If the developer’s Christmas bonus depends on scoring high on a particular benchmark, it is not inconceivable that benchmark somehow would make its way into the training set - directly or indirectly.<p>I don’t think the management would have to directly encourage it as stated in the Chinese text.
152334H大约 1 个月前
VP of AI at <i>FAIR</i>, who is unrelated to the llama&#x2F;genai team.