Chenyan Xiong Research Group at CMU

EmbodiedMidtrain: Bridging the Gap between Vision-Language Models and Vision-Language-Action Models via Mid-training

RePro: Training Language Models to Faithfully Recycle the Web for Pretraining