Audio-Reasoner: Improving Reasoning Capability in Large Audio Language Models
It addresses the overlooked audio modality in multimodal reasoning, providing a domain-specific solution for audio tasks.
The paper tackles the problem of limited reasoning capabilities in large audio language models by introducing Audio-Reasoner, which achieves state-of-the-art performance with improvements such as +25.42% on MMAU-mini and +14.57% on AIR-Bench chat.
Recent advancements in multimodal reasoning have largely overlooked the audio modality. We introduce Audio-Reasoner, a large-scale audio language model for deep reasoning in audio tasks. We meticulously curated a large-scale and diverse multi-task audio dataset with simple annotations. Then, we leverage closed-source models to conduct secondary labeling, QA generation, along with structured COT process. These datasets together form a high-quality reasoning dataset with 1.2 million reasoning-rich samples, which we name CoTA. Following inference scaling principles, we train Audio-Reasoner on CoTA, enabling it to achieve great logical capabilities in audio reasoning. Experiments show state-of-the-art performance across key benchmarks, including MMAU-mini (+25.42%), AIR-Bench chat/foundation(+14.57%/+10.13%), and MELD (+8.01%). Our findings stress the core of structured CoT training in advancing audio reasoning.