Neuron-Level Emotion Control in Speech-Generative Large Audio-Language Models

• URL: http://arxiv.org/abs/2603.17231v1
• Date: Wed, 18 Mar 2026 00:32:47 GMT
• Title: Neuron-Level Emotion Control in Speech-Generative Large Audio-Language Models
• Abstract summary: We present the first neuron-level study of emotion control in large audio-language models (LALMs)<n> compact emotion-sensitive neurons (ESNs) are causally actionable, enabling training-free emotion steering at inference time.<n>Our results establish a mechanistic framework for training-free emotion control in speech generation.
• Score: 15.554739266839187
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: 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.

Related papers

• Emotion Across Speech and Faces: Shared Affective Mechanisms in Multimodal Foundation Models [7.8273212511900985]
  We explore emotion-sensitive neurons (ESNs), sparse decoder neurons selectively associated with emotion categories.<n>Visual ESNs are causally meaningful: deactivating them selectively impairs recognition of the associated facial emotion.<n>Cross-modal interventions reveal bidirectional causal transfer: ESNs identified from one modality produce emotion-specific effects when applied to the other.
  arXiv  Detail & Related papers  (2026-08-17T20:26:22Z)
• A Unified Spoken Language Model with Injected Emotional-Attribution Thinking for Human-like Interaction [50.05919688888947]
  This paper presents a unified spoken language model for emotional intelligence, enhanced by a novel data construction strategy termed Injected Emotional-Attribution Thinking (IEAT)<n>IEAT incorporates user emotional states and their underlying causes into the model's internal reasoning process, enabling emotion-aware reasoning to be internalized rather than treated as explicit supervision.<n> Experiments on the Human-like Spoken Dialogue Systems Challenge (HumDial) Emotional Intelligence benchmark demonstrate that the proposed approach achieves top-ranked performance across emotional trajectory modeling, emotional reasoning, and empathetic response generation.
  arXiv  Detail & Related papers  (2026-01-08T14:07:30Z)
• Discovering and Causally Validating Emotion-Sensitive Neurons in Large Audio-Language Models [8.550786156000461]
  We present the first neuron-level interpretability study of emotion-sensitive neurons (ESNs) in large audio-language models (LALMs)<n>We compare frequency-, entropy-, magnitude-, and contrast-based neuron selectors on multiple emotion recognition benchmarks.<n>Using inference-time interventions, we reveal a consistent emotion-specific signature.
  arXiv  Detail & Related papers  (2026-01-06T15:46:35Z)
• Do LLMs "Feel"? Emotion Circuits Discovery and Control [54.57583855608979]
  We study the internal mechanisms that give rise to emotional expression and in controlling emotions in generated text.<n>This is the first systematic study to uncover and validate emotion circuits in large language models.
  arXiv  Detail & Related papers  (2025-10-13T12:24:24Z)
• EmoCAST: Emotional Talking Portrait via Emotive Text Description [56.42674612728354]
  EmoCAST is a diffusion-based framework for precise text-driven emotional synthesis.<n>In appearance modeling, emotional prompts are integrated through a text-guided decoupled emotive module.<n>EmoCAST achieves state-of-the-art performance in generating realistic, emotionally expressive, and audio-synchronized talking-head videos.
  arXiv  Detail & Related papers  (2025-08-28T10:02:06Z)
• UDDETTS: Unifying Discrete and Dimensional Emotions for Controllable Emotional Text-to-Speech [61.989360995528905]
  We propose UDDETTS, a universal framework unifying discrete and dimensional emotions for controllable emotional TTS.<n>This model introduces the interpretable Arousal-Dominance-Valence (ADV) space for dimensional emotion description and supports emotion control driven by either discrete emotion labels or nonlinearly quantified ADV values.<n>Experiments show that UDDETTS achieves linear emotion control along three interpretable dimensions, and exhibits superior end-to-end emotional speech synthesis capabilities.
  arXiv  Detail & Related papers  (2025-05-15T12:57:19Z)
• From Rational Answers to Emotional Resonance: The Role of Controllable Emotion Generation in Language Models [16.350658746140788]
  Large language models (LLMs) struggle to express emotions in a consistent, controllable, and contextually appropriate manner.<n>We propose a controllable emotion generation framework based on Emotion Vectors (EVs)<n>Our method enables fine-grained, continuous modulation of emotional tone without any additional training or architectural modification.
  arXiv  Detail & Related papers  (2025-02-06T13:38:57Z)
• PROEMO: Prompt-Driven Text-to-Speech Synthesis Based on Emotion and Intensity Control [20.873353104077857]
  We introduce an approach centered on prompt-based emotion control.<n>The proposed architecture incorporates emotion and intensity control across multi-speakers.<n>We leverage large language models (LLMs) to manipulate speech prosody while preserving linguistic content.
  arXiv  Detail & Related papers  (2025-01-10T12:10:30Z)
• EmoKnob: Enhance Voice Cloning with Fine-Grained Emotion Control [7.596581158724187]
  EmoKnob is a framework that allows fine-grained emotion control in speech synthesis with few-shot demonstrative samples of arbitrary emotion. We show that our emotion control framework effectively embeds emotions into speech and surpasses emotion expressiveness of commercial TTS services.
  arXiv  Detail & Related papers  (2024-10-01T01:29:54Z)
• Attention-based Interactive Disentangling Network for Instance-level Emotional Voice Conversion [81.1492897350032]
  Emotional Voice Conversion aims to manipulate a speech according to a given emotion while preserving non-emotion components. We propose an Attention-based Interactive diseNtangling Network (AINN) that leverages instance-wise emotional knowledge for voice conversion.
  arXiv  Detail & Related papers  (2023-12-29T08:06:45Z)
• Emotion Intensity and its Control for Emotional Voice Conversion [77.05097999561298]
  Emotional voice conversion (EVC) seeks to convert the emotional state of an utterance while preserving the linguistic content and speaker identity. In this paper, we aim to explicitly characterize and control the intensity of emotion. We propose to disentangle the speaker style from linguistic content and encode the speaker style into a style embedding in a continuous space that forms the prototype of emotion embedding.
  arXiv  Detail & Related papers  (2022-01-10T02:11:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.