rubric-grader 0.1.0

A Python library for rubric-guided grading with LLMs

rubric-grader

A CLI tool and Python package to grade code submissions using LLM-based rubrics and ensemble code evaluation.

Features

• Scoring modes:
  • A: One-step rubric-based evaluation (default).
  • B: Two-step rubric-based evaluation.
  • C: Ensemble code evaluation.
  • D: AI-O one-shot evaluation.
• Programmatic API via the eval_submissions() function.
• CLI entry point: rubric-grader.
• Example smoke-test script in test/tester.py.

Installation

From PyPI

Install the latest released version from PyPI:

pip install  rubric-grader

From source (editable)

To install the latest development version directly from the GitHub repository, clone the repo and install in editable mode:

git clone https://github.com/arnavthestud/Rubric-Grader.git
cd Rubric-Grader
python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -e .

Usage

Environment Variables

Before running the CLI or using the programmatic API, set your OpenAI API key:

export OPENAI_API_KEY="your_openai_api_key"

CLI

rubric-grader RUBRIC_FILE MODEL_SOLUTION_FILE PROBLEM_STATEMENT_FILE SUBMISSIONS_DIR [OPTIONS]

Example:

rubric-grader test/rubric.txt test/sol.txt test/prob.txt test/sub

Options:

--scoring_type {A,B,C,D}    Scoring mode: A = one-step rubric evaluation (default); B = two-step rubric evaluation; C = ensemble code evaluation; D = AI-O one-shot evaluation
--output_csv OUTPUT_CSV      Path for CSV results (default: results.csv)
--log_file LOG_FILE          Path for log file (default: evaluation.log)
--syntaxMarks SYNTAXMARKS    Maximum syntax marks (default: 5)
--penalty PENALTY            Penalty per syntax error (default: 1)
--ensemble_size ENSEMBLE     Ensemble size for scoring type C (default: 5)
--file_ext FILE_EXT          Submission file extension (default: .txt)
--debug                      Enable debug output

Programmatic API

You can also call the grading function directly from Python:

from llm_grader.code_evaluator import eval_submissions

eval_submissions(
    rubric_filepath='test/rubric.txt',
    model_solution_filepath='test/sol.txt',
    problem_statement_filepath='test/prob.txt',
    submissions_dir='test/sub',
    scoring_type='C',      # or 'B', 'D' (defaults to 'A' for one-step rubric evaluation if omitted)
    output_csv='output/results.csv',
    log_file='output/evaluation.log',
    syntaxMarks=5,
    penalty=1,
    ensemble_size=5,
    debug=False,
    file_ext='.java'
)

Example Test Script

A simple smoke-test script is provided at test/tester.py. Run it with:

python test/tester.py

Inspect test/tester.py to see how it imports eval_submissions() and sets up example file paths.

Scoring Types

• A: One-step rubric evaluation Assigns scores based on a single rubric-based prompting phase.
• B: Two-step rubric evaluation Uses the rubric file to assign scores in two phases (one-step parsing, then detailed rubric criteria).
• C: Ensemble code evaluation Runs an ensemble of LLM queries (default size 5) to judge each submission’s correctness and syntax.
• D: AI-O one-shot evaluation Performs a one-shot evaluation using the AI-O prompt for logical correctness and syntax.

Contributing

Contributions, issues, and feature requests are welcome. Feel free to open a pull request.