[2304.08485] Visual Instruction Tuning
6
公开标注数
6
参与人数
2026-07-18 09:42:20
首次 Whisper
本页的公开 Whisper
划选高亮2026-07-18 16:06:20
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 basically anyone can just build stuff right on arXiv now, that's pretty cool.
划选高亮2026-07-18 13:03:20
原文高亮摘录
“NeurIPS 2023 Oral; project page: this https URL”
Whisper 随想笔记
I wonder if the project page has the actual code or just a demo.
划选高亮2026-07-18 12:54:20
原文高亮摘录
“NeurIPS 2023 Oral; project page: this https URL”
Whisper 随想笔记
Oral at NeurIPS, that's a big deal for this kind of work.
划选高亮2026-07-18 10:00:20
原文高亮摘录
“Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks”
Whisper 随想笔记
This might finally make vision models follow instructions like a champ.
划选高亮2026-07-18 09:51:20
原文高亮摘录
“Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks”
Whisper 随想笔记
I wonder if the 92% on Science QA actually generalizes or just memorizes.
划选高亮2026-07-18 09:42:20
原文高亮摘录
“Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks”
Whisper 随想笔记
Cool, so they just used GPT-4 to make fake data then trained on it, lol.
分享本页 Whisper
短链接
https://domwhisper.com/s/837a7bfad5d9嵌入代码
<iframe src="https://domwhisper.com/embed/837a7bfad5d9" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>