AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

AgentSteerTTS: A Multi-Agent Closed-Loop Framework for Composite-Instruction Text-to-Speech

Bin Kang, Shaoguo Wen, Yang Fan, Shunlong Wu, Junjie Wang, Yulin Li, Junzhi Zhao, Junle Wang, Zhuotao Tian

ICML 2026 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Language-grounded reinforcement agents · 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
add adversarial disentanglement (ablation)AgentSteerTTS · Composite-instruction benchmarkDelta S-SIM: 0.048 -> 0.021Improved
add adversarial disentanglement (ablation)AgentSteerTTS · Composite-instruction benchmarkS-SIM: 0.807 -> 0.841Improved
add adversarial disentanglement (ablation)AgentSteerTTS · Composite-instruction benchmarkCSR: 0.61 -> 0.78Improved
add prototype retrieval (ablation)AgentSteerTTS · Composite-instruction benchmarkE-SIM: 0.910 -> 0.955Improved
add prototype retrieval (ablation)AgentSteerTTS · Composite-instruction benchmarkCSR: 0.49 -> 0.78Improved
add perceptual cropping of retrieved prototype (ablation)AgentSteerTTS · Composite-instruction benchmarkLeakage: 0.21 -> 0.14Improved
add fast loop (ablation)AgentSteerTTS · Composite-instruction benchmarkCSR: 0.70 -> 0.78Improved
add slow loop (ablation)AgentSteerTTS · Composite-instruction benchmarkCSR: 0.67 -> 0.78Improved
add fast loop (ablation)AgentSteerTTS · Composite-instruction benchmarkE-SIM: 0.951 -> 0.955No clear effect
increase prototype library from 25h to 100h (ablation)AgentSteerTTS · Composite-instruction benchmarkE-SIM: 0.912 -> 0.955Improved
increase Fast Loop steps T from 0 to 2 (ablation)AgentSteerTTS · Composite-instruction benchmarkCSR: 0.70 -> 0.78Improved
increase Fast Loop steps T from 3 to 5 (ablation)AgentSteerTTS · Composite-instruction benchmarkCSR: +0.01No clear effect

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