Wei-Teng Hu

Reasoning-Oriented Post-Training and Inference-Time LoRA Rescaling for Audio-Dependent Question Answering

A structured Chain-of-Thought framework is introduced that decomposes the reasoning process into question analysis, question type, audio evidence, and reasoning, and how task-specific LoRA adaptation affects the two backbones is analyzed and inference-time rescaling of trained LoRA adapters is explored.