超越GPT-5？苹果发布AI研究新成果，探索生成UI新方法

Group 1 - Apple's research team has published a study on training artificial intelligence to generate functional user interface code that can compile and align with user prompts regarding functionality and appearance [2] - The study critiques existing reinforcement learning methods based on human feedback, stating they do not align well with designers' workflows [2] - To address this, the researchers proposed a new approach where professional designers provide feedback, sketches, and modifications to the AI-generated interfaces, which are then used to fine-tune the model [2] Group 2 - The best-performing model, Qwen3-Coder, which was fine-tuned using this specific method, outperformed GPT-5 [2]