Jiachen Zhu

transformer architecture (1) contrastive learning (1) mathematical reasoning (1) self-supervised learning (1) multimodal learning (1) model architecture (1) neural network optimization (1) out-of-distribution generalization (1) visual representation learning (1) model training (1) normalization layer (1) retrieval-augmented generation (1) vision encoder (1) layer normalization (1) semantic retrieval (1) process reward model (1) neural network (1) vision-language pretraining (1) reasoning step (1) dynamic tanh (1)