1 paper · 1 filter
Jiuhai Chen, Rifaa Qadri, Yuxin Wen +4
Most public instruction finetuning datasets are relatively small compared to the closed source datasets used to train industry models. To study questions about finetuning at scale,…