AdaRFT: Adaptive Curriculum Reinforcement Finetuning
📢 New extension to verl! We propose an adaptive curriculum learning method for efficient and scalable reinforcement finetuning (RFT) of LLMs — now implemented as an extension to this repo.
Efficient Reinforcement Finetuning via Adaptive Curriculum LearningTaiwei Shi†, Yiyang Wu†, Linxin Song†, Tianyi Zhou▽, Jieyu Zhao††University of Southern California, ▽University of Maryland[Paper]
