Artemis0430/deepseek-v4-pro-opencode-skill-use-data-2487

Deepseek-V4-Pro Opencode Skill-Use data 2487

This dataset was produced by the SKT data pipeline. It contains OpenCode skill-use data integrated with dpsk (DeepSeek V4 Pro), selected from the strict-good trajectory pool for training and research.

Private candidate dataset with three directly viewable Hugging Face splits:

Split	Rows
k1	1,215
k2	776
k3	496

Every row passed the frozen strict selection contract: status=reviewed, response_valid=true, verdict=good, keep_for_training=true, recommended_action=keep, and no safety redaction.

The Data Viewer exposes searchable scalar metadata such as task ID, K, skill IDs, total score, confidence, and digests. Exact heterogeneous trajectory and review objects are stored as valid JSON strings in model_input_json and review_json; parse them with json.loads when loading the dataset:

import json
from datasets import load_dataset

ds = load_dataset("Artemis0430/deepseek-v4-pro-opencode-skill-use-data-2487", token=True)
trajectory = json.loads(ds["k1"][0]["model_input_json"])
review = json.loads(ds["k1"][0]["review_json"])

This repository is private and is not public-rights-cleared. Before changing visibility, complete PII, copyright, third-party skill/license, deduplication, and final diversity/quota review.

Citation

If you use this dataset, please consider citing:

@misc{tan2026sktskillusetrainingscale,
  title={SKT: Skill-Use Training at Scale via Verified Synthetic Data Generation},
  author={Zelin Tan and Yiqun Zhang and Hao Li and Zhiyao Cui and Hejia Geng and Shao Zhang and Hangfan Zhang and Yang Chen and Xiaosong Wang and Lilong Wang and Zhenfei Yin and Shuyue Hu and Chen Zhang and Lei Bai},
  year={2026},
  eprint={2608.02287},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url={https://arxiv.org/abs/2608.02287},
}