AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Sparse Autoencoders for Interpretable Emotion Control in Text-to-Speech

Hongfei Du, Jiacheng Shi, Sidi Lu, Gang Zhou, Ashley Gao

ICML 2026 · Large language models › Language, safety and interpretability · approach: Mechanistic analysis of attention and representations · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
steer 6 selectivity-ranked SAE latents instead of 6 random SAE latents (ablation)SAE-Emotion on IndexTTS2 · Emotion induction, happinessEmo-SIM: 0.813 -> 0.885Improved
use sparse latent steering instead of dense mean-difference global steering (ablation)IndexTTS2 · Emotion induction, anger/happiness/sadnessEmo-SIM: 0.910 -> 0.912 (anger); 0.879 -> 0.885 (happiness); 0.876 -> 0.880 (sadness)Improved
use sparse latent steering instead of global steering for emotion suppression (ablation)IndexTTS2 · Emotion suppression, angerEmo-SIM: 0.915 -> 0.939Improved
use sparse latent steering instead of global steering (ablation)IndexTTS2 · Emotion suppression, WER (anger)WER: 2.6 -> 2.8Hurt
use sparse latent steering instead of global steering at a high steering scale (ablation)IndexTTS2 · Strong steering robustnessMean WER: 2.86% -> 0.57%Improved
select latents by sentence-level selectivity rather than magnitude (ablation)SAE-Emotion · Anger inductionEmo-SIM: 0.822 -> 0.912Improved
select latents by sentence-level selectivity rather than token-level criterion (ablation)SAE-Emotion · Happiness inductionEmo-SIM: 0.811 -> 0.885Improved
increase SAE latent dimension from 4096 to 10,240 (ablation)SAE on IndexTTS2 · SAE trainingemotion interpretabilityHurt
increase number of steered latent features from 1 to 3 to 6 (ablation)SAE-Emotion · Anger and sadness prototype alignmentemotion alignmentImproved
steer one emotion-related latent feature positivelyIndexTTS2 · Paired acoustic analysismean F0 (Hz): 167.99 -> 191.10Improved
steer one emotion-related latent feature positivelyIndexTTS2 · Paired acoustic analysisduration (s): 1.685 -> 1.662No clear effect
increase steering scale alpha from -60 to +60 (ablation)Happiness latent feature · Emotion prototype similaritycosine similarity: 0.77 -> 0.86Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy