StrixAE Agent Orchestrates Audio Enhancement via Multimodal LLMs

StrixAE uses an MLLM controller to coordinate specialized audio models for complex distortion scenarios. It employs Chain-of-Thought supervised fine-tuning and Audio Perception Reinforcement Learning (APRL) to optimize structural coherence and perceptual quality.