DomWhisper

排行榜
登录注册
arxiv.org favicon

[2412.19437] DeepSeek-V3 Technical Report

https://arxiv.org/abs/2412.19437
10
公开标注数
5
参与人数
2026-07-21 09:42:37
首次 Whisper

本页的公开 Whisper

arxiv.org favicon
arxiv.org
划选高亮2026-07-23 13:27:24
原文高亮摘录
“arXivLabs is a framework that allows collaborators to develop and share new arXiv features”
Whisper 随想笔记
Sounds like a nice idea, but will it actually lead to useful features or just clutter?
arxiv.org favicon
arxiv.org
划选高亮2026-07-23 13:18:24
原文高亮摘录
“arXivLabs is a framework that allows collaborators to develop and share new arXiv features”
Whisper 随想笔记
Interesting, but who actually gets to be a 'collaborator' here?
arxiv.org favicon
arxiv.org
划选高亮2026-07-23 10:24:24
原文高亮摘录
“a strong Mixture-of-Experts (MoE) language model with 671B total parameters”
Whisper 随想笔记
2.788M H800 hours for that size? My cluster would melt trying that.
arxiv.org favicon
arxiv.org
划选高亮2026-07-23 10:15:24
原文高亮摘录
“a strong Mixture-of-Experts (MoE) language model with 671B total parameters”
Whisper 随想笔记
Wait, so MoE means most of those 671B params are just sitting idle?
arxiv.org favicon
arxiv.org
划选高亮2026-07-23 10:06:24
原文高亮摘录
“a strong Mixture-of-Experts (MoE) language model with 671B total parameters”
Whisper 随想笔记
Honestly, 671B feels like overkill when you only use 37B per token.
arxiv.org favicon
arxiv.org
划选高亮2026-07-21 13:03:37
原文高亮摘录
“arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website”
Whisper 随想笔记
Cool, but I bet the review process is gonna be a pain to get anything in.
arxiv.org favicon
arxiv.org
划选高亮2026-07-21 12:54:37
原文高亮摘录
“arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website”
Whisper 随想笔记
So they're basically letting random devs build stuff right into the site?
arxiv.org favicon
arxiv.org
划选高亮2026-07-21 10:00:37
原文高亮摘录
“a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token”
Whisper 随想笔记
Training on 14.8T tokens with that few GPU hours is insane, my last model needed more for a fraction of that.
arxiv.org favicon
arxiv.org
划选高亮2026-07-21 09:51:37
原文高亮摘录
“a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token”
Whisper 随想笔记
Mixture-of-Experts keeps getting bigger, wonder if it's just a trick to inflate parameter counts.
arxiv.org favicon
arxiv.org
划选高亮2026-07-21 09:42:37
原文高亮摘录
“a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token”
Whisper 随想笔记
671B params but only 37B active, that's like having a huge team but only a few show up to work.

分享本页 Whisper

分享到 X
短链接
https://domwhisper.com/s/de539966b9f5
嵌入代码
<iframe src="https://domwhisper.com/embed/de539966b9f5" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>

看看大家在 arxiv.org 上讨论了什么

安装 DomWhisper，浏览网页时实时查看 whisper，也可以加入讨论。

获取插件