Improving Audio Captioning Models with Fine-Grained Audio Features, Text Embedding Supervision, and LLM Mix-Up Augmentation.