datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
haotian_data-GPS-AR-Lopti-master
Do Not Let Low-Probability Tokens Over-Dominate in RL for LLMs
Authors: Zhihe Yang$^{1}$, Xufang Luo$^{2*}$, Zilong Wang$^2$, Dongqi Han$^2$, Zhiyuan He$^2$, Dongsheng Li$^2$, Yunjian Xu$^{1*}$,
($^*$ for corresponding authors)
The Chinese University of Hong Kong, Hong Kong SAR, China
Microsoft Research Asia, Shanghai, China
Introduction
In this study, we identify a critical yet underexplored issue in RL training: low-probability tokens disproportionately… See the full description on the dataset page: https://huggingface.co/datasets/happynew111/haotian_data-GPS-AR-Lopti-master.gps-image-datasetgps-image-dataset_newgps