AI Paper Atlas

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

DualSpeechLM: Towards Unified Speech Understanding and Generation via Dual Speech Token Modeling with Large Language Models

Yuanyuan Wang, Dongchao Yang, Yiwen Shao, Hangting Chen, Jiankun Zhao, Zhiyong Wu, Helen Meng, Xixin Wu

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Speech and audio language models · 3 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
use USToken instead of HuBERT as input token (ablation)DualSpeechLM (Phi3.5-3B) · LibriSpeech test-clean ASRWER: 5.56 -> 4.22Improved
use USToken instead of HuBERT as input token (ablation)DualSpeechLM (Phi3.5-3B) · TTS LibriTTS cleanWER: 20.80 -> 9.25Improved
use USToken instead of HuBERT as input token (ablation)DualSpeechLM (Vicuna-7B) · TTS LibriTTS cleanWER: 13.58 -> 12.63Improved
add AcousticGPT dual-token output (vs single-token baseline)HuBERT-input Phi3.5-3B · TTS LibriTTS cleanWER: 21.72 -> 20.80Improved
add understanding-driven loss in USTokenizer (ablation)DualSpeechLM · SQABLEU4: 37.67 -> 44.38Improved
add understanding-driven loss in USTokenizer (ablation)DualSpeechLM · TTS LibriTTS cleanWER: 8.59 -> 9.25No clear effect
add reconstruction loss in USTokenizer (ablation)DualSpeechLM · TTS LibriTTS cleanWER: 52.50 -> 9.25Improved
add reconstruction loss in USTokenizer (ablation)DualSpeechLM · T2ST Es2EnBLEU4: 32.43 -> 26.77Hurt
add semantic supervision loss (ablation)DualSpeechLM · TTS LibriTTS cleanWER: 167.56 -> 9.25Improved
add semantic supervision loss (ablation)DualSpeechLM · SQABLEU4: 43.86 -> 44.38No clear effect
add Chain of Condition (ablation)DualSpeechLM · VC VCTKWER: 13.06 -> 10.16Improved
switch backbone from Phi3.5-3B to Vicuna-7B (ablation)DualSpeechLM (USToken) · LibriSpeech test-clean ASRWER: 4.22 -> 4.15No 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