---
language:
- en
license: mit
size_categories:
- 1K<n<10K
task_categories:
- time-series-forecasting
- tabular-classification
- tabular-regression
tags:
- time-series
- foundation-models
- probing
- benchmark
- evaluation
- synthetic
- baseline-controlled
- neurips-2026
pretty_name: TSFMI-Synthetic
configs:
- config_name: trend
  data_files:
    - split: train
      path: trend/train.parquet
    - split: validation
      path: trend/val.parquet
    - split: test
      path: trend/test.parquet
- config_name: seasonality
  data_files:
    - split: train
      path: seasonality/train.parquet
    - split: validation
      path: seasonality/val.parquet
    - split: test
      path: seasonality/test.parquet
- config_name: frequency
  data_files:
    - split: train
      path: frequency/train.parquet
    - split: validation
      path: frequency/val.parquet
    - split: test
      path: frequency/test.parquet
- config_name: stationarity
  data_files:
    - split: train
      path: stationarity/train.parquet
    - split: validation
      path: stationarity/val.parquet
    - split: test
      path: stationarity/test.parquet
- config_name: anomaly
  data_files:
    - split: train
      path: anomaly/train.parquet
    - split: validation
      path: anomaly/val.parquet
    - split: test
      path: anomaly/test.parquet
- config_name: change_point
  data_files:
    - split: train
      path: change_point/train.parquet
    - split: validation
      path: change_point/val.parquet
    - split: test
      path: change_point/test.parquet
- config_name: trend_hard
  data_files:
    - split: train
      path: trend_hard/train.parquet
    - split: validation
      path: trend_hard/val.parquet
    - split: test
      path: trend_hard/test.parquet
- config_name: frequency_hard
  data_files:
    - split: train
      path: frequency_hard/train.parquet
    - split: validation
      path: frequency_hard/val.parquet
    - split: test
      path: frequency_hard/test.parquet
- config_name: stationarity_hard
  data_files:
    - split: train
      path: stationarity_hard/train.parquet
    - split: validation
      path: stationarity_hard/val.parquet
    - split: test
      path: stationarity_hard/test.parquet
- config_name: anomaly_hard
  data_files:
    - split: train
      path: anomaly_hard/train.parquet
    - split: validation
      path: anomaly_hard/val.parquet
    - split: test
      path: anomaly_hard/test.parquet
- config_name: change_point_hard
  data_files:
    - split: train
      path: change_point_hard/train.parquet
    - split: validation
      path: change_point_hard/val.parquet
    - split: test
      path: change_point_hard/test.parquet
---

# TSFMI-Synthetic

Synthetic time-series datasets with **mathematically exact ground-truth labels**
for the TSFMI baseline-controlled probing protocol.

> Companion data for the NeurIPS 2026 Evaluations & Datasets Track submission
> *"TSFMI: A Baseline-Controlled Evaluation Protocol for Time-Series Foundation
> Model Representations."* The code (anonymous) lives at
> https://anonymous.4open.science/r/TSFMI.

## Why this dataset exists

Probing time-series foundation models (TSFMs) is hard because high probe
accuracy may reflect probe capacity rather than encoded knowledge. TSFMI
addresses this with **explicit non-model controls** (hand-crafted features,
raw signal, random projection, ROCKET) evaluated under the same canonical
60/20/20 split, the same sklearn estimator, the same 5 seeds, and the same
bootstrap CI as the model probe. The synthetic datasets in this repository
provide six standard temporal properties **plus five hard variants** under
which TSFM representation claims can be evaluated against the controls.

## Configurations (11 tasks)

| Config | Type | Classes / Range | Description |
|---|---|---|---|
| `trend` | classification | 3 (up / down / flat) | linear slope ± noise |
| `seasonality` | regression | period ∈ {8, 16, 32, 64} | sinusoid + noise; label = period |
| `frequency` | classification | 8 frequency bins | discrete sinusoidal frequency bands |
| `stationarity` | classification | 2 (stationary / non-stationary) | Gaussian noise vs. random walk |
| `anomaly` | classification | 2 (normal / anomaly) | single ±5σ point spike — **kurtosis is a sufficient statistic** |
| `change_point` | classification | 2 (none / has-CP) | mean+variance shift at midpoint |
| `*_hard` (5) | classification | varies | structured background + subtle anomalies / weak slopes / overlapping harmonics / variance drift / mild CP |

Each config has 1000 sequences of length 512, partitioned 60/20/20
(train/val/test = 600/200/200). All data is generated deterministically with
`seed=42` (data) and `split_seed=0` (partition), matching the canonical
artefacts in the paper.

## Quick start

```python
from datasets import load_dataset

ds = load_dataset("EvalData/TSFMI", "anomaly")
print(ds)
# DatasetDict({
#   train: 600, validation: 200, test: 200
# })
print(ds["train"][0]["sequence"][:8], ds["train"][0]["label"])
```

Each row contains:

- `sequence`: list[float] of length 512 (the time-series window)
- `label`: int (classification) or float (seasonality regression)
- `seed`: int (data-generation seed; always 42 here)
- `split_seed`: int (split seed used to partition this row; 0 for the canonical artefacts)

## Reproduction recipe

The same 60/20/20 + 5-seed + bootstrap protocol used in the paper is
implemented in [the GitHub repository](https://anonymous.4open.science/r/TSFMI):

```bash
curl -L -o TSFMI.zip "https://anonymous.4open.science/api/repo/TSFMI/zip"
unzip TSFMI.zip && cd TSFMI
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt && pip install -e ".[dev]"

# CPU smoke test (<10 min) — reproduces the headline HC anomaly = 0.858 cell
make smoke

# Full canonical pipeline (~48 A100-hours)
make extract-representations
make reproduce-all
```

## Headline result reproducible from this dataset

Under the canonical 60/20/20 protocol with sklearn `LogisticRegression` and a
bootstrap 95% CI over 5 seeds:

- An 8-D **hand-crafted feature vector** reaches **0.858** test accuracy on
  the canonical `anomaly` configuration.
- A **single kurtosis feature** alone reaches **0.859**.
- A **single max-absolute-magnitude feature** alone reaches **0.907**.
- The best of seven pre-trained TSFMs reaches **0.753** (TimesFM); all other
  TSFMs (MOMENT, Chronos, PatchTST, GPT4TS, Timer, Moirai) score 0.50–0.73.

This **inversion** is what motivates the baseline-controlled discipline of
TSFMI and is documented in §3.3 and Appendix A.8 of the paper.

## License

MIT. The synthetic data is fully procedurally generated; there is no
human-derived or scraped content. Real-world datasets used elsewhere by the
TSFMI evaluation pipeline (ETTh1, Weather, Electricity, Traffic, Exchange
Rate, UCR) are **not redistributed here** and remain under their respective
original licenses; see the GitHub `LICENSE` file for details.

## Citation

```bibtex
@misc{tsfmi2026,
  title = {{TSFMI}: A Baseline-Controlled Evaluation Protocol for Time-Series Foundation Model Representations},
  author = {Anonymous Authors},
  howpublished = {Anonymous submission to the NeurIPS 2026 Evaluations \& Datasets Track},
  year = {2026},
  url = {https://anonymous.4open.science/r/TSFMI}
}
```

## Responsible AI notes

- **Data collection**: synthetic procedural generation, no human subjects, no scraping.
- **Limitations**: each generator instantiates one statistically simple test bed; the canonical anomaly task is kurtosis-trivial by design (`realistic_anomaly` variant in the GitHub repo bounds this).
- **Biases**: none — purely synthetic with no demographic content.
- **Personal/sensitive info**: none.
- **Use cases**: probing TSFM internal representations, calibrating new probing protocols against simple non-model baselines.
- **Misuse cases**: not intended as a production anomaly detector; do not interpret the canonical anomaly task as evidence that any model "detects anomalies" in any operational sense.
- **Synthetic data indicator**: 100% synthetic.

The full Croissant 1.0 + RAI metadata is provided as `CROISSANT.json` in this
repository.