1. **Open-Source Models Only**  
   You are only allowed to use **open-weight models** in your solution. The use of proprietary, paid, or closed-source models is strictly prohibited. Violation of this rule will result in **disqualification**.

2. **Code Sharing Requirement (Top 5)**  
   To promote transparency and learning, participants who finish in the **Top 5** on the private leaderboard are required to publicly share their **full training and inference code** that can reproduce their final score.  
   - You’ll have **2 days after the competition ends** to submit your reproducible code.  
   - If code is not submitted or the result cannot be reasonably reproduced, we may need to **remove the entry from the final rankings**.
   - Preferably the teams should write a small article or blog explaining the logic behind their model choice and how they got there.

3. **No Duplicate Accounts**  
   The use of **duplicate or multiple accounts** is strictly prohibited. It can be easily discovered by the host. This will be actively monitored and violations will result in **immediate disqualification**.

4. **Prize Eligibility**  
   Prizes will be awarded **only to participants residing in Saudi Arabia**. If you are not a resident of Saudi Arabia, you are welcome to participate for learning and leaderboard placement, but you will not be eligible for monetary prizes.

5. **Manual Annotation of Test Set Is Prohibited**  
   Manually labeling or annotating the test set (e.g., by reading and assigning topics by hand) is strictly **not allowed**. Any indication of manual labeling will result in **disqualification** This also includes closed sourced automatic LLM labeling.