# MI-Fuse: Label Fusion for Unsupervised Domain Adaptation with Closed-Source Large-Audio Language Model

arXiv: https://arxiv.org/abs/2509.20706 · HTML: https://mlsanity.com/paper/2509.20706/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 21 of its 30 references that are in the map.
- First posted on arXiv 25 Sep 2025.

## Builds on (most cited first, 21 of 21)

- [Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning](https://mlsanity.com/paper/1506.02142/) (2015)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://mlsanity.com/paper/2110.13900/) (2021)
- [Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation](https://mlsanity.com/paper/2002.08546/) (2020)
- [Qwen2.5-Omni Technical Report](https://mlsanity.com/paper/2503.20215/) (2025)
- [A Survey on Deep Semi-supervised Learning](https://mlsanity.com/paper/2103.00550/) (2021)
- [Exploiting the Intrinsic Neighborhood Structure for Source-free Domain Adaptation](https://mlsanity.com/paper/2110.04202/) (2021)
- [Source Data-absent Unsupervised Domain Adaptation through Hypothesis Transfer and Labeling Transfer](https://mlsanity.com/paper/2012.07297/) (2020)
- [Source-Free Domain Adaptation for Semantic Segmentation](https://mlsanity.com/paper/2103.16372/) (2021)
- [Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks](https://mlsanity.com/paper/2411.05361/) (2024)
- [C-SFDA: A Curriculum Learning Aided Self-Training Framework for Efficient Source Free Domain Adaptation](https://mlsanity.com/paper/2303.17132/) (2023)
- [DeSTA2.5-Audio: Toward General-Purpose Large Audio Language Model with Self-Generated Cross-Modal Alignment](https://mlsanity.com/paper/2507.02768/) (2025)
- [Towards audio language modeling -- an overview](https://mlsanity.com/paper/2402.13236/) (2024)
- [Emo-bias: A Large Scale Evaluation of Social Bias on Speech Emotion Recognition](https://mlsanity.com/paper/2406.05065/) (2024)
- [Speech Emotion Recognition using Supervised Deep Recurrent System for Mental Health Monitoring](https://mlsanity.com/paper/2208.12812/) (2022)
- [Feature Alignment by Uncertainty and Self-Training for Source-Free Unsupervised Domain Adaptation](https://mlsanity.com/paper/2208.14888/) (2022)
- [Self Supervised Adversarial Domain Adaptation for Cross-Corpus and Cross-Language Speech Emotion Recognition](https://mlsanity.com/paper/2204.08625/) (2022)
- [Mitigating Subgroup Disparities in Multi-Label Speech Emotion Recognition: A Pseudo-Labeling and Unsupervised Learning Approach](https://mlsanity.com/paper/2505.14449/) (2025)
- [EMO-Debias: Benchmarking Gender Debiasing Techniques in Multi-Label Speech Emotion Recognition](https://mlsanity.com/paper/2506.04652/) (2025)
- [Dataset-Distillation Generative Model for Speech Emotion Recognition](https://mlsanity.com/paper/2406.02963/) (2024)
- [Utilizing Speech Emotion Recognition and Recommender Systems for Negative Emotion Handling in Therapy Chatbots](https://mlsanity.com/paper/2311.11116/) (2023)

## Abstract (arXiv)

Large audio-language models (LALMs) show strong zero-shot ability on speech tasks, suggesting promise for speech emotion recognition (SER). However, SER in real-world deployments often fails under domain mismatch, where source data are unavailable and powerful LALMs are accessible only through an API. We ask: given only unlabeled target-domain audio and an API-only LALM, can a student model be adapted to outperform the LALM in the target domain? To this end, we propose MI-Fuse, a denoised label fusion framework that supplements the LALM with a source-domain trained SER classifier as an auxiliary teacher. The framework draws multiple stochastic predictions from both teachers, weights their mean distributions by mutual-information-based uncertainty, and stabilizes training with an exponential moving average teacher. Experiments across three public emotion datasets and six cross-domain transfers show consistent gains, with the student surpassing the LALM and outperforming the strongest baseline by 3.9%. This approach strengthens emotion-aware speech systems without sharing source data, enabling realistic adaptation.