CodeReviewBench / src /about.py
Alex
zalupa1
982b341
raw
history blame
1.96 kB
"""
About page content for CodeReview Leaderboard
"""
TITLE = "🏆 CodeReview Leaderboard"
INTRODUCTION_TEXT = """
# CodeReview Leaderboard
A comprehensive benchmark for evaluating code review generation models across multiple programming languages and comment types.
## Overview
This leaderboard tracks the performance of various models on code review tasks, providing insights into:
- **Programming Language Performance**: How well models perform across different programming languages
- **Comment Language Support**: Effectiveness in generating reviews in different natural languages
- **Taxonomy Categories**: Performance across different types of code review feedback
## Metrics
- **BLEU**: Measures similarity between generated and reference reviews
- **Pass@1/5/10**: Percentage of reviews that pass quality checks in 1, 5, or 10 attempts
- **Multi-dimensional Quality Scores**: Detailed evaluation across 10 quality dimensions
## Features
✨ **Filter by Programming Language**: View results for specific programming languages (Python, JavaScript, Java, etc.)
✨ **Comment Language Support**: Filter by the natural language of code comments
✨ **Taxonomy Categories**: Browse results by review type (bug detection, style, performance, etc.)
✨ **IP-based Submissions**: Secure submission system with IP tracking
✨ **Dark Theme**: Modern, eye-friendly interface
"""
SUBMISSION_GUIDELINES = """
## Submission Guidelines
1. **Model Requirements**: Submit results for at least 100 test cases
2. **Format**: Provide scores in the specified format ranges
3. **Reproducibility**: Include model details and evaluation setup
4. **Quality Metrics**: Rate your model across all 10 quality dimensions
5. **Metadata**: Specify programming language, comment language, and taxonomy focus
"""
CONTACT_INFO = """
## Contact & Support
For questions, issues, or contributions, please reach out through our repository or contact the maintainers.
"""