sparc-visualization 0.1.3

No project description provided

Visual SPaRC: Visual Puzzle Representations for the SPaRC benchmark

📦 PyPI Package | 🤗 SPaRC Dataset

Overview

SPaRC (sparc-puzzle) provides a comprehensive framework for evaluating language models on spatial reasoning tasks inspired by "The Witness" puzzle game. This package (sparc-visualization) includes various visual puzzle representations and an improved visual reasoning prompt to complement the functionality of the sparc-puzzle package.

Installation

Install the package from PyPI:

pip install sparc-visualization

Or install from source:

git clone https://github.com/flowun/sparc-visualization.git
cd sparc-visualization
pip install -e .

Example Usage

from sparc_visualization.plot import get_puzzle_image
from sparc_visualization.prompt import generate_prompt
from sparc.validation import extract_solution_path, validate_solution, analyze_path
from datasets import load_dataset

# Load the dataset
dataset = load_dataset("lkaesberg/SPaRC", "all", split="test")
puzzle = dataset[0]

# Generate prompt and image, e.g.
plot_type = "path_cell_annotated"
prompt_type = "prompt_engineering"

b64_image = get_puzzle_image(puzzle, plot_type=plot_type, base_64_image=True)
text_prompt = generate_prompt(puzzle, plot_type, prompt_type)

# create message array and payload for an LLM
puzzle_prompt = [
        {
            "role": "user",
            "content": [
                {
                    "type": "image_url",
                    "image_url": {
                        "url": f"data:image/png;base64,{b64_image}"
                    }
                },
                {
                    "type": "text",
                    "text": text_prompt
                }
            ]
        }
    ]

payload = {
    "model": your_model_name,
    "messages": puzzle_prompt,
    "temperature": temperature,
    "max_tokens": max_tokens,
}

# Your model generates a response
model_response = "... model response with path coordinates ..."

# Analysis using sparc-puzzle package

# Extract the path from model response
extracted_path = extract_solution_path(model_response, puzzle)
# Returns: [{"x": 0, "y": 2}, {"x": 0, "y": 1}, ...]

# Validate against ground truth
is_correct = validate_solution(extracted_path, puzzle)
# Returns: True/False

# Get detailed analysis
analysis = analyze_path(extracted_path, puzzle)
# Returns: {
#   "starts_at_start_ends_at_exit": True,
#   "connected_line": True,
#   "non_intersecting_line": True,
#   "no_rule_crossing": True,
#   "fully_valid_path": True
# }

Core Functions

get_puzzle_image(
    puzzle: Dict,
    plot_type: str = "original",
    base_64_image: bool = False,
    path: List[Tuple[int, int]] | List[Dict[str, int]] | None = None,
    show_plot: bool = False,
    save_to_disk: bool = False,
    save_dir: str = ".",
    save_filename: str = "puzzle_image.png",
) -> PIL.Image.Image | str

• Renders a SPaRC puzzle into one of the supported visual representations (plot_type values).
• To get a base64-encoded string of the image suitable for LLM input, set base_64_image=True.
• To overlay a path, specify path either as a list of (x, y) tuples or a list of {"x": int, "y": int} cells. The full solution path can be displayed by passing puzzle["solutions"][0]["path"] to this argument.
• Besides returning the image object or base64 string, the function has options to display the image (show_plot=True) or save it to disk (save_to_disk=True).

generate_prompt(
    puzzle: Dict,
    plot_type: str = "original",
    prompt_type: str = "prompt_engineering",
) -> str

Generates the text prompt (based on prompt_type values) that should be paired with the rendered image for an LLM call. Available prompt_type values are:

• "default_tr": visual prompt from the SPaRC paper (with textual coordinates)
• "default_no_tr": visual prompt from the SPaRC paper (however with textual coordinates removed)
• "prompt_engineering": improved vision-only prompt with prompt engineering (no textual coordinates) puzzle and plot_type should be provided the same as in get_puzzle_image to allow small prompt adjustments.

Available Puzzle Representations (plot_type)

original	start_end_marked	coordinate_grid
coordinate_grid_and_start_end_marked	path_cell_annotated	text
low_contrast	low_contrast_and_path_cell_annotated	low_resolution
low_resolution_and_path_cell_annotated	rotated	rotated_and_path_cell_annotated
black_frame	black_frame_and_path_cell_annotated

Contributing

Contributions are welcome! Please feel free to submit pull requests or open issues.

Citation

If you use SPaRC in your research, please cite:

@article{kaesberg2025sparc,
  title     = {SPaRC: A Spatial Pathfinding Reasoning Challenge},
  author    = {Kaesberg, Lars Benedikt and Wahle, Jan Philip and Ruas, Terry and Gipp, Bela},
  year      = {2025},
  url       = {https://arxiv.org/abs/2505.16686}
}

License

This project is licensed under the MIT License - see the LICENSE file for details.

Links

• 🌐 Website: sparc.gipplab.org
• 📚 Dataset: Hugging Face
• 🐛 Issues: GitHub Issues
• 📖 Documentation: GitHub Repository