Confucius4-R2T2 released as stable-prefix streaming ASR model
Original title: Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription
R2T2 maintains stable prefixes in real-time transcription without modifying earlier outputs while open-sourced by Netease Youdao
We present Confucius4-R2T2, a stable-prefix streaming ASR model obtained by post-training Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio while freezing its pre-trained audio encoder and projector. The adaptation is based on the concept of Longest Stable Prefix (LSP), a new notion we introduce. Learning from LSP labels converts a full-context model into a stable-prefix streaming one without any architectural change. R2T2 brings streaming recognition close to the base model's full-context quality while delivering stable text with low emission lag. Its gains extend beyond the post-training languages.
- Post-trained on 1,500 hours of Chinese and English audio
- Freezes pre-trained audio encoder and projector
- Never modifies earlier output
- Open-source at Hugging Face and GitHub
- Vendors
- Netease Youdao
- AI models
- Confucius4-R2T2
From the original article
1 min read at huggingface.co
Streaming speech recognition is increasingly consumed by machines, such as translators and voice agents, that act on each token as it appears. A token that is later revised costs wasted work or a wrong action. In this work, we present Confucius4-R2T2 (short for Real Real-Time Transcription), a stable-prefix streaming ASR model that never modifies its earlier output while maintaining high accuracy and low latency.
By Daniel Huang, Qi He, Shaoxiong Duan · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.