### 1. Competition Overview * **Start Date:** 20 February 2026. * **Team Size:** Strictly 1 (Individual participation only). * **Task:** Solve 16 independent levels of increasing complexity using Reinforcement Learning. --- ### 2. Core Participation Rules * **Learning-Based Solutions Only:** Participants must design an agent that learns to solve each level through training. * **Mandatory Learning Curves:** The agent must produce a visible learning curve (e.g., reward vs. episode, loss vs. epoch). * **No Hardcoding:** Hardcoded solutions, heuristics, or direct analytical solvers (mathematical shortcuts) are strictly prohibited. * **AI Tools:** The use of Large Language Models (LLMs) and other AI tools is strongly encouraged. --- ### 3. Submission & Manual Verification To ensure the integrity of the leaderboard, winning solutions will undergo a **manual code audit**. Each participant must submit two notebooks: * **** Must define the environment (Gymnasium, PyGame, etc.), implement the training loop, show a clear learning curve, and save the trained weights. * **** Must load the trained weights and run the agent without further training to demonstrate successful task completion. --- ### 4. Ranking and Tie-Breaking * **Level Completion:** Ranking is determined by the highest-numbered level completed. * **Submission Timing:** If multiple participants complete the same highest level, the participant who submitted earlier ranks higher. --- ### 5. Prohibited Conduct * **Cheating:** Any attempt to bypass the learning process or use hardcoded logic will result in disqualification. * **Multiple Accounts:** You cannot sign up to Kaggle from multiple accounts and therefore you cannot submit from multiple accounts. * **Private Sharing:** Outside of public Kaggle forums, private sharing of code or solutions is not allowed. --- ### 6. Disclaimer ENTRY IN THIS COMPETITION CONSTITUTES YOUR ACCEPTANCE OF THESE OFFICIAL COMPETITION RULES. ANYONE FOUND VIOLATING ANY OF THESE RULES WILL BE DISQUALIFIED.