# Steering Follows Geometry, Not Labels: Emotion Directions in a Full-Duplex Speech Model

arXiv: https://arxiv.org/abs/2610.08887 · HTML: https://mlsanity.com/paper/2610.08887/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Activation steering methods, one of 13 directions in Mechanistic interpretability.
- Builds on 15 of its 18 references that are in the map.
- First posted on arXiv 6 Oct 2026.

## Placed in

- Mechanistic interpretability / [Activation steering methods](https://mlsanity.com/mechinterp/activation-steering-methods/)

## Builds on (most cited first, 15 of 15)

- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [Steering Language Models With Activation Engineering](https://mlsanity.com/paper/2308.10248/) (2023)
- [Moshi: a speech-text foundation model for real-time dialogue](https://mlsanity.com/paper/2410.00037/) (2024)
- [emotion2vec: Self-Supervised Pre-Training for Speech Emotion Representation](https://mlsanity.com/paper/2312.15185/) (2023)
- [PersonaPlex: Voice and Role Control for Full Duplex Conversational Speech Models](https://mlsanity.com/paper/2602.06053/) (2026)
- [Understanding (Un)Reliability of Steering Vectors in Language Models](https://mlsanity.com/paper/2505.22637/) (2025)
- [EmoSphere-TTS: Emotional Style and Intensity Modeling via Spherical Emotion Vector for Controllable Emotional Text-to-Speech](https://mlsanity.com/paper/2406.07803/) (2024)
- [EmoSteer-TTS: Fine-Grained and Training-Free Emotion-Controllable Text-to-Speech via Activation Steering](https://mlsanity.com/paper/2508.03543/) (2025)
- [Controlling Emotion in Text-to-Speech with Natural Language Prompts](https://mlsanity.com/paper/2406.06406/) (2024)
- [Understanding Unreliability of Steering Vectors in Language Models: Geometric Predictors and the Limits of Linear Approximations](https://mlsanity.com/paper/2602.17881/) (2026)
- [Overcoming State Inertia in Full-Duplex Spoken Language Models via Activation Steering](https://mlsanity.com/paper/2606.11386/) (2026)
- [JoyAI-Talker: Full-Duplex Speech Interactive Large Model Built for Empathetic Voice Agents](https://mlsanity.com/paper/2608.01119/) (2026)
- [F-Actor: Controllable Conversational Behaviour in Full-Duplex Models](https://mlsanity.com/paper/2601.11329/) (2026)
- [Sparse Autoencoders for Interpretable Emotion Control in Text-to-Speech](https://mlsanity.com/paper/2606.01479/) (2026)
- [A Geometric Perspective on Composable Emotion Steering in Text-to-Speech Models](https://mlsanity.com/paper/2607.00946/) (2026)

## Abstract (arXiv)

Full-duplex voice agents need to modulate emotion and delivery during real-time conversations, when de-escalating a complaint, carrying urgency in dispatch, softening a clinical result. Emotion and delivery control is well studied for TTS and turn based models through prompt-conditioned synthesis, reference-conditioned synthesis and activation steering; PersonaPlex controls identity in a duplex model but not affect. We study emotion steering in Moshi, a fully open sourced full-duplex speech language model, across four emotions, using mean-difference activation steering, which costs only a few vector additions per frame and no retraining. We show that emotion is linearly decodable from Moshi's residual stream, but activation steering is only partially achievable, and unevenly so; as happy, angry and surprise steer towards a shared direction while sad is distinctly steerable. We also show that the shared component across the three emotions cannot simply be projected away from all the emotions equally.