🚀 Featured Project
Multimodal Understanding Foundation Model for Visual Quality Reasoning
â–² Objective: Develop a vision-language foundation model for professional visual quality reasoning, with image quality assessment serving as a high-standard evaluation scenario.
â–² Research Scope: (1) Hundred-GPU-scale VLM Training & Evaluation Infra; (2) Multi-task Joint Training; (3) Cross-model Hyperparameter Transfer (MuP); (4) Release Model Selection.
â–² Applications: (1) Reliable Assessment for Generative Algorithms; (2) Reward Model (RM) for Post-training.