Generates synthetic datasets for word-search letter-grid tracing. The agent must locate a hidden word inside a grid of letters and trace its path — testing pattern matching across multiple letters in straight or non-straight directions.
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-13 |
| Task | Word Search Path Trace |
| Category | Constraint Satisfaction Puzzles |
| 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-13_wordsearch_path_data-generator.git
cd Multi-13_wordsearch_path_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 grid of letters containing a hidden target word.
[Rules]
1. The target word can be hidden horizontally, vertically, or diagonally.
2. The letters of the word must be adjacent to each other in a straight line.
[Task] Generate a video showing the process of finding the hidden word. Animate a continuous highlight over the correct sequence of letters that spells the target word.
 |
 |
 |
| Initial Frame Letter grid with target word listed |
Animation Path traced cell-by-cell to reveal word |
Final Frame Hidden word fully highlighted |
Locate and trace a hidden target word inside a procedurally generated grid of letters, where the word can be embedded in straight directions (horizontal, vertical, diagonal) and possibly bent.
- Grid: NxN cells filled with random letters; one path encodes the target word.
- Target word: Drawn from an English wordlist; declared in the prompt.
- Directions: Horizontal, vertical, diagonal — any straight or path-traced sequence.
- Distractors: All other cells contain randomly sampled letters that may share prefixes with the target.
- Solver: Deterministic search over all valid letter placements; ground-truth path is recorded at generation time.
- Multi-cell sequential pattern matching: Recognizing the word requires aligning multiple consecutive cell observations.
- Direction ambiguity: Many false leads share initial letters; the agent must validate the full sequence.
- Visual trace verification: Each step of the trace is rendered, providing intermediate correctness signal.
- Long-horizon symbolic reasoning: Identification spans the full word length (typically 5–10 cells).
data/questions/Multi-13_wordsearch_path_data-generator_task/Multi-13_wordsearch_path_data-generator_00000000/
├── first_frame.png # Letter grid with target word listed
├── final_frame.png # Hidden word path highlighted
├── prompt.txt # Task instruction
├── ground_truth.mp4 # Animation of cell-by-cell trace
└── 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
word-search letter-grid pattern-matching path-tracing csp multi-step-reasoning symbolic
Part of the 36-Task Long-Horizon Multi-Step Reasoning Benchmark.