CoolFace
Datasetpublic

Dayanand314Krishna/cross_rulings_hts_dataset_for_tariffs

CROSS Rulings HTS Dataset for Tariff Classification Maintained by Flexify.AI Inc. as part of the ATLAS trade intelligence research program. Paper: ATLAS: Benchmarking and Adapting LLMs for Global Trade via Harmonized Tariff Code Classification Project Page: https://tariffpro.flexify.ai/ This dataset is constructed from the U.S. Customs and Border Protection (CBP) Rulings Online Search System (CROSS).It contains rulings where importers sought clarification on the correct… See the full description on the dataset page: https://huggingface.co/datasets/Dayanand314Krishna/cross_rulings_hts_dataset_for_tariffs.

sourceHugging Faceapache-2.0updated 8mo agoView on Hugging Face
0likes182downloads
Dataset Card

CROSS Rulings HTS Dataset for Tariff Classification

Maintained by Flexify.AI Inc. as part of the ATLAS trade intelligence research program.

Paper: ATLAS: Benchmarking and Adapting LLMs for Global Trade via Harmonized Tariff Code Classification Project Page: https://tariffpro.flexify.ai/

This dataset is constructed from the U.S. Customs and Border Protection (CBP) Rulings Online Search System (CROSS). It contains rulings where importers sought clarification on the correct Harmonized Tariff Schedule (HTS) codes.

Contents

  • —18,731 rulings across 2,992 unique HTS codes
  • —Train / validation / test splits (18,254 / 200 / 200)
  • —Structured in an LLM-trainable format:
  • —Product description
  • —Reasoning path
  • —Correct HTS code

Task

The dataset supports HTS code prediction as:

  • —10-digit U.S.-specific HTS code classification
  • —6-digit globally harmonized HS code classification

This makes it suitable for:

  • —Text classification
  • —Question answering
  • —Reasoning with structured outputs

Benchmark Results

We evaluate LLaMA-3.3-70B, GPT-5-Thinking, Gemini-2.5-Pro-Thinking, and our fine-tuned Atlas model on 200 held-out CROSS rulings.

Model10-digit Accuracy6-digit AccuracyAvg. Digits Correct
GPT-5-Thinking25.0%55.5%5.61
Gemini-2.5-Pro-Thinking13.5%31.0%2.92
DeepSeek-R1 (05/28)2.5%26.5%3.24
GPT-OSS-120B1.5%8.0%2.58
LLaMA-3.3-70B (baseline)2.1%20.7%3.31
Atlas (Fine-tuned LLaMA)40.0%57.5%6.30

📖 Citation

If you use this dataset, please cite the following:

bibtex
@dataset{flexify_cross_rulings_hts,
  title={CROSS Rulings HTS Dataset for Tariff Classification},
  author={{Flexify.AI Inc.}},
  year={2025},
  url={https://huggingface.co/datasets/flexifyai/cross_rulings_hts_dataset_for_tariffs},
  publisher={Flexify.AI Inc.},
  howpublished={\url{https://www.flexify.ai}}
}

If you reference the associated research, please cite the paper separately:
@misc{atlas2025benchmarking,
  title={ATLAS: Benchmarking and Adapting LLMs for Global Trade via Harmonized Tariff Code Classification}, 
  author={Siva Devarakonda and others},
  year={2025},
  eprint={2509.18400},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url={https://arxiv.org/abs/2509.18400}
}

## Maintainer

- **Siva Devarakonda** — Founder & CEO, Flexify.AI Inc.


## License & Attribution

This dataset is released under the Apache License 2.0.

Any use of this dataset or derivatives (including trained models, benchmarks, publications, or downstream applications) must include attribution as follows:

> “CROSS Rulings HTS Dataset for Tariff Classification by Flexify.AI Inc. (https://www.flexify.ai)”

Please also include a link to the original dataset:
https://huggingface.co/datasets/flexifyai/cross_rulings_hts_dataset_for_tariffs

and clearly indicate if any modifications were made.
Please ensure compliance with U.S. Customs regulations when using for downstream tasks.

## Acknowledgments
- Data derived from the [U.S. Customs Rulings Online Search System (CROSS)](https://rulings.cbp.gov/)  
- Harmonized Tariff Schedule reference: [USITC](https://hts.usitc.gov/)