[2210.11416] Scaling Instruction-Finetuned Language Models
6
公开标注数
6
参与人数
2026-07-18 10:05:56
首次 Whisper
本页的公开 Whisper
划选高亮2026-07-18 16:29:56
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
I mean, cool but who actually uses arXivLabs for real work?
划选高亮2026-07-18 13:26:56
原文高亮摘录
“Code, Data and Media Associated with this Article”
Whisper 随想笔记
Just code and data? No pretrained weights? That's kinda disappointing honestly.
划选高亮2026-07-18 13:17:56
原文高亮摘录
“Code, Data and Media Associated with this Article”
Whisper 随想笔记
Wish more papers listed their datasets like this, saves hours of hunting.
划选高亮2026-07-18 10:23:56
原文高亮摘录
“Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks.”
Whisper 随想笔记
I wonder if the chain-of-thought data is what really pushed it over the edge.
划选高亮2026-07-18 10:14:56
原文高亮摘录
“Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks.”
Whisper 随想笔记
Honestly this matches my experience, teaching models with examples just works.
划选高亮2026-07-18 10:05:56
原文高亮摘录
“Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks.”
Whisper 随想笔记
So the trick is just more tasks and bigger models? Seems too simple tbh.
分享本页 Whisper
短链接
https://domwhisper.com/s/e019a1695aea嵌入代码
<iframe src="https://domwhisper.com/embed/e019a1695aea" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>