| --- |
| 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 |
|
|
| [](https://arxiv.org/abs/2601.23161v1) |
| [](https://huggingface.co/zhoujiaming777/DIFFA-2) |
| [](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. |
