Ziheng Qin

model compression (1) model quantization (1) continual learning (1) catastrophic forgetting (1) binary classification (1) deepfake detection (1) neural network optimization (1) model training (1) vision-language model (1) weight averaging (1) zero-shot transfer (1) edge computing (1) low-bit quantization (1) feature distillation (1) edge device inference (1) forgery feature extraction (1) click-through rate prediction (1) large batch training (1) embedding scaling (1) column-wise clipping (1)