##  Competition Rules

###  General Rules

1. **One Account Per Participant**
   Each participant must have only one Kaggle account. Teams may be formed, but duplicate accounts for an individual are prohibited.
2. **Team Formation**
   Participants can compete individually or in teams of up to 5 members. Team formation and merging must be completed before the team merger deadline.
3. **Competition Timeline**
   All submissions must be made before the competition deadline. Late submissions will not be accepted under any circumstances.
4. **Submission Limit**
   Participants may make up to **2 submissions per day** and retain **their best single score** on the public leaderboard for final evaluation.
5. **Final Evaluation**
   The final ranking will be based on a private leaderboard calculated using a hidden portion of the test data. It may differ from the public leaderboard.



###  Code and Data Rules

6. **Code Sharing**
   Public code sharing (e.g., via Kaggle notebooks, GitHub, forums) is **not allowed during the active phase** of the competition. Public sharing will only be permitted after the **Code Freeze Date (TBD)**.

7. **External Data**
   You may use external data if:

   * It is freely and publicly available
   * You provide a verifiable URL/source
   * It is uploaded to Kaggle Datasets and clearly documented in your notebook

   Use of **private or proprietary data** is strictly prohibited.

8. **Original Work**
   All submissions must be the original work of the participant or team. Use of pre-existing solutions without attribution or plagiarism will result in disqualification.

9. **Automated Tools**
   Use of automated hyperparameter tuning tools (e.g., AutoML) is allowed unless otherwise stated, provided they comply with external data and compute rules.

---

###  Technical Rules

10. **Submission Format**
    Submissions must be in the exact format provided in `sample_submission.csv`. Submissions not matching the format will be rejected.
11. **Model Reproducibility**
    Finalists may be required to submit code and documentation sufficient to reproduce their final results, including data preprocessing and training scripts.
12. **Environment Compliance**
    Your final model must be trainable and executable using reasonable hardware and within a reasonable time (e.g., <12 hours on a standard Kaggle GPU instance).


###  Disqualification Conditions

* Failure to comply with these rules
* Use of unfair or unethical practices
* Sharing or leaking test labels
* Training on test data or manipulating the leaderboard



### Notes

* All decisions by the competition organizers are final.
* These rules may be updated during the competition. All participants will be notified of major changes.