The authors built ELLSA 2 around a dialogue backbone and an autoregressive, chunk-wise synthesizer.
A dialogue model coordinates speech, gestures and expressions continuously
ELLSA 2 generates spoken and non-verbal responses during both speaking and listening, with template guidance for gestures carrying explicit meaning.
Chinese Tech
Siyin Wang · Wenyi Yu · Terumi Chiba · Xianzhao Chen · Xiaohai Tian · Jun Zhang · +2 more
Tsinghua University · ByteDance
Research Digest··2 min read
Wang and colleagues present an end-to-end model for full-duplex interaction, meaning it can listen, speak and produce visible reactions concurrently rather than waiting for fixed conversational turns.
Why this paper
From ByteDance and Tsinghua University
In one line
An end-to-end full-duplex dialogue model coordinates speech, gestures and expressions during speaking and listening, improving semantic gesture quality.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§