Generates synthetic datasets for Conway's Game of Life simulation. The agent must evolve a 2D grid of cells through several generations by applying birth/survival/death rules to all cells simultaneously — the canonical example of long-horizon massively-parallel deterministic simulation.
Each sample pairs a task (first frame + prompt describing what needs to happen) with its ground truth solution (final frame showing the result + video demonstrating how to achieve it). This structure enables both model evaluation and training.
| Property | Value |
|---|---|
| Task ID | Multi-24 |
| Task | Conway's Game of Life |
| Category | Algorithmic Execution |
| Resolution | 1024×1024 px |
| FPS | 16 fps |
| Duration | varies |
| Output | PNG images + MP4 video |
# 1. Clone the repository
git clone https://github.com/VBVR-DataFactory/Multi-24_conways_game_of_life_data-generator.git
cd Multi-24_conways_game_of_life_data-generator
# 2. Create and activate virtual environment
python3 -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# 3. Install dependencies
pip install --upgrade pip
pip install -r requirements.txt
pip install -e .# Generate 50 samples
python examples/generate.py --num-samples 50
# Reproducible generation with seed
python examples/generate.py --num-samples 50 --seed 42
# Custom output directory
python examples/generate.py --num-samples 100 --output data/my_dataset
# Without videos (faster, images only)
python examples/generate.py --num-samples 50 --no-videos| Argument | Description |
|---|---|
--num-samples |
Number of tasks to generate (required) |
--output |
Output directory (default: data/questions) |
--seed |
Random seed for reproducibility |
--no-videos |
Skip video generation (images only) |
[Scenario] The image shows a 2D grid representing the initial state of Conway's Game of Life. Live cells are filled with green color, and dead cells are light gray.
[Rules]
1. Any live cell with fewer than 2 or more than 3 live neighbors dies.
2. Any live cell with 2 or 3 live neighbors lives on to the next generation.
3. Any dead cell with exactly 3 live neighbors becomes a live cell.
[Task] Generate a video simulating the grid evolution generation by generation. Animate the cell states changing according to the rules for exactly 3 generations. Hold the final frame to clearly show the exact board state.
 |
 |
 |
| Initial Frame Generation 0 board state |
Animation Generation-by-generation evolution |
Final Frame Exact board state after N generations |
Evolve a 2D grid of live/dead cells through a fixed number of generations using Conway's three rules, where every cell's next state depends on the current state of all 8 neighbors (synchronous update).
- Grid: 10×10 cells (live = filled, dead = empty).
- Initial population: Random subset of cells alive at generation 0.
- Rules:
- Underpopulation: Live cell with <2 live neighbors dies.
- Survival: Live cell with 2 or 3 live neighbors lives.
- Reproduction: Dead cell with exactly 3 live neighbors becomes alive.
- Generations: Typically 3 (configurable).
- Update: All cells update simultaneously based on the previous generation — no sequential cell-by-cell mutation.
- Massively parallel deterministic update: Every cell's next-state computation is independent but global state evolves in lock-step — a quintessential parallel algorithm.
- Long-horizon emergent dynamics: Patterns can stabilize, oscillate, glide, or extinguish — small initial differences yield wildly different N-step futures.
- No shortcut from local pattern matching: Each generation requires reasoning about all cells' neighborhoods at once.
- Frame-perfect intermediate states: Each generation transition is rendered, providing per-generation evaluation hooks.
data/questions/Multi-24_conways_game_of_life_data-generator_task/Multi-24_conways_game_of_life_data-generator_00000000/
├── first_frame.png # Generation 0 board
├── final_frame.png # Generation N board
├── prompt.txt # Task instruction with rules
├── ground_truth.mp4 # Animation of generation evolution
└── question_metadata.json # Standardized VBVR task metadata
File specifications:
- Images: 1024×1024 PNG format
- Video: MP4 format, 16 fps, H.264 + yuv420p
- Metadata: VBVR canonical schema with
task_id,vbvr_task_code,media,parameters
game-of-life cellular-automaton synchronous-update algorithmic-execution simulation emergent-dynamics long-horizon multi-step-reasoning
Part of the 36-Task Long-Horizon Multi-Step Reasoning Benchmark.