Tracing a Sparse Emotion-Control Circuit in LLM-Based Text-to-Speech

Excerpt

arXiv:2610.05080v1 Announce Type: cross Abstract: LLM-based text-to-speech (TTS) models can generate emotionally expressive speech, but how reference emotion is routed through the model and realized in decoded speech remains unclear. We introduce two emotion-sensitive metrics for matched neutral and emotional syntheses---a codec trajectory score and a late residual direction score---and use them to score activation-patching interventions. Under controlled matched-reference conditions, this analy