The authors built TimeInteract around three components: a streaming encoder that represents both recent changes and longer-term history, a controller that decides whether a response should be triggered, and an inference system that separates this controller from language generation.
A model monitors live time series while generating responses concurrently
TimeInteract learns when to respond and continues processing incoming measurements while composing its answer.
Research Lab
Sheng Pan · Yongli Gu · Yiqing Guo · Warren Jin · Bo Du · Shirui Pan · +1 more
Griffith University · CSIRO
Research Digest··2 min read
Pan et al.
Why this paper
From CSIRO and Griffith University
In one line
TimeInteract lets time-series language models stream observations and generate responses concurrently, outperforming prior models by up to 23.92 points.
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.
§