Hugging Face's logo

Datasets:
AINativeOps
/
AINativeBench

Dataset card Files
xet
 Community
AINativeBench / data /processed /RQ1 /BookWriter-MCP /README.md
王子睿
restructure + add files
8c10cf2
preview code
|
Raw
History Blame Contribute Delete
1.12 kB

BookWriter-MCP (RQ1)

This folder contains analysis scripts and reference configurations used to evaluate BookWriter-MCP experiment results.

Files

  • evaluate_trajectory.py

    • Evaluates execution trajectories extracted from execution_path.md.
    • Outputs CSV results.
    • Markdown output is intentionally disabled.

  • reference_trajectory.yaml

    • Defines the reference (ground-truth) trajectory and evaluation configuration.

  • analyze_retry_patterns.py

    • Analyzes retry patterns from execution_path.md across models.
    • Outputs JSON/CSV summaries.

  • evaluate_success.py

    • Computes success rate based on produced chapter/book .md artifacts in each session folder.
    • Outputs JSON/CSV summaries.

Usage

Run scripts from this directory, for example:

python3 evaluate_trajectory.py --config reference_trajectory.yaml --format csv --output evaluation_results.csv
python3 analyze_retry_patterns.py
python3 evaluate_success.py

Notes

  • These scripts analyze existing experiment outputs under /Users/wzr/TOSEM-2025/RESULTS.
  • The scripts do not generate any Markdown reports.