ChorusTIC

ChorusTIC is a classification-native foundation model for training-free univariate and multivariate time series classification. It uses labeled context examples to predict query labels without fitting a target-specific classifier or updating model parameters.

Paper · Code · Official benchmark results

Benchmark and evaluation

ChorusTIC is evaluated in the TSC-FM time series classification foundation model benchmark. Use the time series classification leaderboard to compare matching Standard and low-shot settings, and read the evaluation protocol before interpreting aggregate accuracy, rank and coverage.

Model description

ChorusTIC combines Random Subchannel Slot Concatenation with a shared dual-axis encoder to model temporal and cross-channel interactions across heterogeneous channel configurations. Context-derived feature calibration and leakage-protected in-context learning are then used to infer query labels.

The released repository contains ChorusTIC.ckpt. Model loading, preprocessing and UCR/UEA evaluation code are available in the official GitHub repository.

Intended use

• Research on training-free time series classification.
• In-context classification of univariate and multivariate sequences.
• Reproduction and extension of the accompanying paper’s experiments.

Limitations

• Predictions require labeled context examples with adequate class coverage.
• Performance outside the evaluated data domains and channel configurations is not guaranteed.
• This checkpoint is a research artifact and should be validated independently before high-stakes use.

Citation

@article{fang2026chorustic,
  title   = {ChorusTIC: Training-Free Multivariate Time Series Classification via Chorus In-Context Learning},
  author  = {Fang, Juntao and Xie, Shifeng and Cai, Ruichu and Zheng, Shengji and Li, Zijian and Zhang, Keli and Pan, Lujia and Palpanas, Themis and Hao, Zhifeng},
  journal = {arXiv preprint arXiv:2608.24033},
  year    = {2026}
}