Neuron-Level Emotion Control in Speech-Generative Large Audio-Language Models
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866908896580337664 |
|---|---|
| author | Zhao, Xiutian Ulgen, Ismail Rasim Koehn, Philipp Schuller, Björn Sisman, Berrak |
| author_facet | Zhao, Xiutian Ulgen, Ismail Rasim Koehn, Philipp Schuller, Björn Sisman, Berrak |
| contents | Large audio-language models (LALMs) can produce expressive speech, yet reliable emotion control remains elusive: conversions often miss the target affect and may degrade linguistic fidelity through refusals, hallucinations, or paraphrase. We present, to our knowledge, the first neuron-level study of emotion control in speech-generative LALMs and demonstrate that compact emotion-sensitive neurons (ESNs) are causally actionable, enabling training-free emotion steering at inference time. ESNs are identified via success-filtered activation aggregation enforcing both emotion realization and content preservation. Across three LALMs (Qwen2.5-Omni-7B, MiniCPM-o 4.5, Kimi-Audio), ESN interventions yield emotion-specific gains that generalize to unseen speakers and are supported by automatic and human evaluation. Controllability depends on selector design, mask sparsity, filtering, and intervention strength. Our results establish a mechanistic framework for training-free emotion control in speech generation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_17231 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Neuron-Level Emotion Control in Speech-Generative Large Audio-Language Models Zhao, Xiutian Ulgen, Ismail Rasim Koehn, Philipp Schuller, Björn Sisman, Berrak Computation and Language Audio and Speech Processing Large audio-language models (LALMs) can produce expressive speech, yet reliable emotion control remains elusive: conversions often miss the target affect and may degrade linguistic fidelity through refusals, hallucinations, or paraphrase. We present, to our knowledge, the first neuron-level study of emotion control in speech-generative LALMs and demonstrate that compact emotion-sensitive neurons (ESNs) are causally actionable, enabling training-free emotion steering at inference time. ESNs are identified via success-filtered activation aggregation enforcing both emotion realization and content preservation. Across three LALMs (Qwen2.5-Omni-7B, MiniCPM-o 4.5, Kimi-Audio), ESN interventions yield emotion-specific gains that generalize to unseen speakers and are supported by automatic and human evaluation. Controllability depends on selector design, mask sparsity, filtering, and intervention strength. Our results establish a mechanistic framework for training-free emotion control in speech generation. |
| title | Neuron-Level Emotion Control in Speech-Generative Large Audio-Language Models |
| topic | Computation and Language Audio and Speech Processing |
| url | https://arxiv.org/abs/2603.17231 |