Checklist HAIO 2025 Summer National Final · Task 2
Catacomb Adventure
English title: Katakomba Kaland
Guide a heroine out of a grid catacomb in a custom Gymnasium environment with search algorithms, Monte Carlo control and SARSA.
The task
A heroine is trapped in a catacomb with pits, traps and zombies and must reach the exit door. The organisers provide a custom Gymnasium environment (CreepyCatacombs-v0, package creepy-catacombs-s1) whose map depends on the seed, with optional zombies (n_zombies) moving randomly or towards the player. Actions are 0 up, 1 right, 2 down, 3 left. Rewards: -1 per step, -5 for bumping into a wall, -100 for falling into a pit or meeting a zombie (episode ends), 0 for reaching the exit (episode ends).
Task 1 (5 points): a hard-coded action sequence that reaches the exit on the seed-2025 map without zombies, printing the final position and total reward. Task 2 (20 points): a path-finding algorithm (e.g. backtracking or BFS) that avoids walls and obstacles. Task 3 (10 points): evaluate that solution with 1, 5 and 9 randomly moving zombies over 10 episodes each (seeds 2025 + episode index) and plot the rewards on one figure. Task 4 (15 points): extend the path finder to avoid zombies as well.
Task 5 (20 points): complete an every-visit Monte Carlo control implementation with an ε-greedy policy. Task 6 (5 points): visualise the learned state values, Q-values and optimal path with the environment's render_values, render_q_values and render_optimal_path helpers. Task 7 (20 points): complete a SARSA implementation. Task 8 (5 points): repeat the visualisations for SARSA. Task 9 (+15 bonus points): plot moving-average rewards of Monte Carlo and SARSA together and rerun Monte Carlo with two randomly moving zombies, describing the differences.
Abridged by SOTA from the official materials. The official statement has the exact rules, and it wins wherever this summary differs.
In English
Some of this task's files were published only in Hungarian. SOTA translated that file into English on 16 September 2026.
Read the task description (one-pager) in English
Catacomb Adventure
English translation by SOTA – AI Community of the Hungarian original. Licensed CC BY-NC-SA 4.0, like the original. Organisers who would like this translation removed can email [email protected].
Magyar MI Diákolimpia (Hungarian AI Olympiad)
Catacomb Adventure (Katakomba Kaland)
Task description
24 May 2025
1. Catacomb Kata
Kátai Kata (better known as Catacomb Kata – Katakomba Kata) decided one Monday afternoon to go looking for diamonds. Not because she is a romantic type, but because she had heard that diamonds make the best pickaxe, one that can break even obsidian. And with a tool like that you can mine anything: for example, meaning in life after the school-leaving exams, or at least a way out of a catacomb.
Things, however (not very surprisingly), did not go according to plan. Kata got lost. Instead, she found herself in a damp labyrinth of angular walls, where her only hope lies not in maps but in algorithms.
Your task: help Kata get out by trying to steer her towards the exit with various reinforcement learning methods. For this, a custom-built gymnasium environment is at your disposal, which can be found here:
During the adventure we will guide you through eight algorithmic challenges, large and small, which you can solve in a Jupyter Notebook (Google Colab or locally). You will need neither a spade nor a torch. It will help, though, if you are not scared of a value function.
2. Scoring
- Exploring the
gymnasiumenvironment – [5 points] - Building a search algorithm – [20 points]
- Running the algorithm and visualising the path – [10 points]
- Modifying the algorithm – [15 points]
- Implementing Monte Carlo Control – [20 points]
- Running Monte Carlo and visualisation – [5 points]
- Completing the SARSA algorithm – [20 points]
- Running SARSA and visualisation – [5 points]
- Visualising the aggregated results and learning curves – [15 points, optional]
Maximum score: 100 points
Optional bonus points available: 15 points
3. Technical information
To solve the task, installing the gymnasium environment and the attached Notebook file are sufficient. The list of required libraries and the installation instructions can be found at the beginning of the notebook.
After solving the task, regardless of whether you worked in Google Colab or locally, you must upload the .ipynb file to the CMS system for the corresponding task. No other file needs to be uploaded.
4. Useful resources
Translated by SOTA. The Hungarian original is the official version and wins wherever the two differ. Original by the Hungarian AI Olympiad (ELTE Faculty of Informatics), licensed CC BY-NC-SA 4.0; this is a translation of the task one-pager, and the official English notebook of the task is linked on this page. This translation is shared under CC BY-NC-SA 4.0, the licence of the original. If you organise this olympiad and would like the translation removed, email [email protected] and we will take it down.
At a glance
- You get
- The creepy-catacombs-s1 package (version 0.1.4 in the notebook) and the task notebook with helper code; no dataset.
- You submit
- The completed .ipynb notebook uploaded to the CMS.
- Scoring
- 100 points plus 15 optional bonus points (5 + 20 + 10 + 15 + 20 + 5 + 20 + 5, bonus 15); graded by the organisers.
- Rules
- Work only on the provided lab machines (Windows); own laptops are not allowed. Google Colab is recommended; a local Python environment is also provided.
- At most one NVIDIA T4 GPU; stronger GPUs (e.g. A100, V100) lead to disqualification.
- General internet access, but code-completion and LLM services (e.g. GitHub Copilot, ChatGPT, Claude) are forbidden; the free tier of Gemini 2.5 Flash integrated in the Google ecosystem is the only exception.
- Publicly available documentation, articles and books may be used during the practical part; audio-visual material is forbidden; communication is forbidden.
- Mandatory screen recording with OBS Studio for the whole contest.
- Format
- Summer National Final (Nyári Országos Döntő), 24 May 2025, on site at ELTE, Budapest. Second (practical) part: two hours of programming after a 15-minute briefing; four 100-point tasks (CV, ML, NLP, RL). Solutions (.ipynb) are uploaded to the CMS.