AuTAgent: A Reinforcement Learning Framework for Tool-Augmented Audio Reasoning

Siqian Tong1,2, Xuan Li1,2, Yiwei Wang3, Baolong Bi2,4, Yujun Cai5,
Shenghua Liu2,4, Yuchen He2,4, Chengpeng Hao1,2
1Institute of Acoustics, CAS   2University of Chinese Academy of Sciences   3University of California, Merced
4Institute of Computing Technology, CAS   5The University of Queensland

Abstract

Large Audio Language Models (LALMs) excel at perception but struggle with complex reasoning requiring precise acoustic measurements. While external tools can extract fine-grained features like exact tempo or pitch, effective integration remains challenging: naively using all tools causes information overload, while prompt-based selection fails to assess context-dependent utility.

To address this, we propose AuTAgent (Audio Tool Agent), a reinforcement learning framework that learns when and which tools to invoke. By employing a sparse-feedback training strategy with a novel Differential Reward mechanism, the agent learns to filter out irrelevant tools and invokes external assistance only when it yields a net performance gain over the base model.

Experimental results confirm that AuTAgent complements the representation bottleneck of LALMs by providing verifiable acoustic evidence. It improves accuracy by 4.20% / 6.20% and 9.80% / 8.00% for open-source and closed-source backbones on the MMAU Test-mini and the MMAR benchmarks, respectively. In addition, further experiments demonstrate exceptional transferability. We highlight the complementary role of external tools in augmenting audio model reasoning.

Method Overview

AuTAgent Framework

The overview of AuTAgent framework. The agent learns to select appropriate tools via reinforcement learning to augment LALM reasoning.

BibTeX

@inproceedings{tong2026autagent,
  title={AuTAgent: A Reinforcement Learning Framework for Tool-Augmented Audio Reasoning},
  author={Tong, Siqian and Li, Xuan and Wang, Yiwei and Bi, Baolong and Cai, Yujun and Liu, Shenghua and He, Yuchen and Hao, Chengpeng},
  booktitle={International Conference on Machine Learning},
  year={2026}
}