Analyze model performance across training stages
Display benchmark evaluation data for LLMs
Generate captions for images