TimeBraid: Unifying Time Series and Language for Understanding and Forecasting
TimeBraid pairs pretrained language and time-series models inside one shared representation space.
The paper says the system uses interleaved global residual attention layers to align the two model types.
It is trained with a unified prompting setup, stabilized joint training, 2.2M curated series-text pairs, and 4.9M instruction-tuning samples.
The authors report competitive results across time-series perception, understanding, reasoning, and forecasting benchmarks.
Source: HF Daily Papers' note
The paper says the system uses interleaved global residual attention layers to align the two model types.
It is trained with a unified prompting setup, stabilized joint training, 2.2M curated series-text pairs, and 4.9M instruction-tuning samples.
The authors report competitive results across time-series perception, understanding, reasoning, and forecasting benchmarks.
Source: HF Daily Papers' note
score 5