[1910.10683] Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
5
公开标注数
5
参与人数
2026-07-21 10:04:57
首次 Whisper
本页的公开 Whisper
划选高亮2026-07-21 13:25:57
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
Hope it stays open and doesn't get too bureaucratic.
划选高亮2026-07-21 13:16:57
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 cool way to get community-built features.
划选高亮2026-07-21 10:22:57
原文高亮摘录
“Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task”
Whisper 随想笔记
I tried this approach once, but my corpus was way smaller. Still got decent results though.
划选高亮2026-07-21 10:13:57
原文高亮摘录
“Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task”
Whisper 随想笔记
Pre-training on a huge corpus then fine-tuning sounds simple, but the details matter a lot.
划选高亮2026-07-21 10:04:57
原文高亮摘录
“Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task”
Whisper 随想笔记
This is basically how every NLP model works now, but they actually made it systematic.
分享本页 Whisper
短链接
https://domwhisper.com/s/887da5f02141嵌入代码
<iframe src="https://domwhisper.com/embed/887da5f02141" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>