AI Paper Atlas

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

UniSS: Unified Expressive Speech-to-Speech Translation with Your Voice

Sitong Cheng, Bianweizhen, Xinsheng Wang, Ruibin Yuan, Jianyi Chen, Shunshun Yin, Yike Guo, Wei Xue

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Speech and audio language models · 10 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 Phase 3 refinement training (ablation)UniSS base (Phase 1+2) · CVSS-T EN-ZHSpeech-BLEU: 29.38 -> 30.28 (+0.90)Improved
add Phase 3 refinement training (ablation)UniSS base (Phase 1+2) · CVSS-T ZH-ENSpeech-BLEU: 21.55 -> 23.61 (+2.06)Improved
add Phase 1 speech-text alignment (ablation)UniSS · CVSS-T EN-ZHSpeech-BLEU: 22.20 -> 29.38 (-7.18 without)Improved
add Phase 1 speech-text alignment (ablation)UniSS · CVSS-T ZH-ENSpeech-BLEU: 11.40 -> 21.55 (-10.15 without)Improved
use GLM-4 linguistic tokenizer instead of BiCodec semantic tokens (ablation)UniSS · CVSS-T EN-ZHSpeech-BLEU: -15.01Improved
use GLM-4 linguistic tokenizer instead of BiCodec semantic tokens (ablation)UniSS · CVSS-T ZH-ENSpeech-BLEU: -8.73Improved
add cross-modal chain-of-thought (intermediate text) instead of direct S2ST (ablation)UniSS · CVSS-T EN-ZHSpeech-BLEU: -14.94Improved
add cross-modal chain-of-thought (intermediate text) instead of direct S2ST (ablation)UniSS · CVSS-T ZH-ENSpeech-BLEU: -14.40Improved
switch from Quality mode to Performance modeUniSS 1.5B · CVSS-TSpeed vs Speech-BLEU: 1.07x speedup with 1.84 point reduction in Speech-BLEUMixed
shrink backbone to Qwen2.5-0.5B (UniSS-Small, Performance mode)UniSS · CVSS-TSpeedup: 1.25x speedup, Speech-BLEU 25.68Mixed
use UniSS (Q) instead of Seamless-Exexpressive S2ST · CVSS-TSLC 0.2 / duration consistency: 44% and 67% improvementImproved
use UniSS (Q) instead of Seamless-Exexpressive S2ST · Expressive emotion datasetEmotion similarity MOS: 3.56 -> 4.51Improved

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