polym-eval-report-generator

To generate an Elo ranking report from SQLite model evaluation data, you need a local EVAL_DB_PATH SQLite database and the generate_report.py script to compute Elo scores and render Markdown evaluation sections.

Yes, Markdown evaluation reports support filtering by specific report_id and model name, allowing you to limit the output to precisely scope win-rate analytics and dimension breakdowns for targeted model assessments.

Yes, the report generator supports Chinese output for image and video model comparisons, allowing you to render Elo summaries, win-rate analytics, and scenario breakdowns in either Chinese or English.

You can save generated evaluation reports to a specified file path to export results for sharing. The tool writes the computed Markdown report content directly to the designated file location upon request.

Yes, a local EVAL_DB_PATH SQLite database is required to compute win-rate analysis. The generator accesses this local database directly to aggregate comparison data and produce the evaluation reports.

A Markdown evaluation report for image and video model comparisons includes Elo rankings, win-rate summaries, and dimension or scenario breakdowns to provide a comprehensive overview of model performance.