Algorithm Benchmark Suite v2.0: Synthetic Dataset for Circular Bias Detection

<h1>Algorithm Benchmark Suite v2.0</h1>

<p>A comprehensive benchmark dataset for evaluating circular bias detection algorithms in AI evaluation contexts.</p>

<h2>Dataset Contents</h2>

<p>This archive contains <strong>10 CSV files</strong> with synthetic evaluation data across diverse scenarios:</p>

<ul><li><strong>algorithm_benchmark_suite.csv</strong> - Main benchmark (20 records, 7 fields)</li><li><strong>scenario_*.csv</strong> (9 files) - Controlled experiments with varying:<ul><li>Noise levels (0.05-0.2)</li><li>Bias intensities (0.0-0.9)</li><li>Temporal configurations (8-20 periods)</li><li>Algorithm counts (3-6 algorithms)</li></ul></li></ul>

<h2>Data Schema</h2>

<p>Each record includes standardized fields:</p><ul><li><code>time_period</code>: Evaluation period ID (integer)</li><li><code>algorithm</code>: Algorithm identifier (string)</li><li><code>performance</code>: Performance metric [0-1] (float)</li><li><code>constraint_compute</code>: Computational constraint (float)</li><li><code>constraint_memory</code>: Memory constraint in GB (float)</li><li><code>constraint_dataset_size</code>: Training dataset size (integer)</li><li><code>evaluation_protocol</code>: Protocol version (string)</li></ul>

<h2>Use Cases</h2><ul><li>Benchmarking bias detection algorithms</li><li>Testing evaluation methodology robustness</li><li>Research on algorithmic fairness in temporal domains</li><li>Validation of circular reasoning detection frameworks</li></ul>

<h2>Reproducibility</h2>

<p>All data is fully reproducible using the included generation script:<br><code>python data/generate_benchmark_data.py --validate</code></p>

<h2>Associated Software</h2>

<p>Framework implementation: <a href="https://github.com/hongping-zh/circular-bias-detection">https://github.com/hongping-zh/circular-bias-detection</a></p>

<h2>Documentation</h2><ul><li><strong>Data Dictionary</strong>: Comprehensive field specifications</li><li><strong>JSON Schema</strong>: Formal validation rules</li><li><strong>README</strong>: Usage examples and guidelines</li><li><strong>CHANGELOG</strong>: Version history and updates</li></ul>

<h2>Citation</h2><pre>Zhang, H. (2024). Algorithm Benchmark Suite v2.0 [Data set]. Zenodo. https://doi.org/10.5281/zenodo.17201032</pre>