README.md

---
language:
- en
base_model:
- GSAI-ML/LLaDA-8B-Instruct
tags:
- audio
- speech
- music
---
# DIFFA-2: A Practical Diffusion Large Language Model for General Audio Understanding

[![arXiv](https://img.shields.io/badge/Paper-arXiv-red.svg)](https://arxiv.org/abs/2601.23161v1)
[![🤗 Hugging Face](https://img.shields.io/badge/🤗Hugging%20Face-DIFFA-FFEB3B)](https://huggingface.co/zhoujiaming777/DIFFA-2)
[![GitHub](https://img.shields.io/badge/Github-DIFFA-blue)](https://github.com/NKU-HLT/DIFFA)

 In this paper, We introduce DIFFA-2, a practical diffusion-based LALM for general audio understanding. DIFFA-2 upgrades the speech encoder, employs dual semantic and acoustic adapters, and is trained with a four-stage curriculum that combines semantic and acoustic alignment, large-scale supervised fine-tuning, and variance-reduced preference optimization, using only fully open-source corpora. Experiments on MMSU, MMAU, and MMAR show that DIFFA-2 consistently improves over DIFFA and is competitive to strong AR LALMs under practical training budgets, supporting diffusion-based modeling is a viable backbone for large-scale audio understanding.

 We have open-sourced the checkpoints for stage 1 and stage 4. The files in the root directory of the repository are for stage4, and stage1 is located in the stage1 folder.