Powerful Teachers Matter: Text-Guided Multi-view Knowledge Distillation with Visual Prior Enhancement Synthetic Mixed Training: Scaling Parametric Knowledge Acquisition Beyond RAG A Deep Dive into Scaling RL for Code Generation with Synthetic Data and Curricula CAPTCHA Solving for Native GUI Agents: Automated Reasoning-Action Data Generation and Self-Corrective Training Scaling and quantization of large-scale foundation model enables resource-efficient predictions in network biology