RQ1

This directory contains scripts and aggregated artifacts for RQ1 evaluation.

Structure

Each subfolder corresponds to one task/project and orchestration variant:

*-A2A: A2A runs
*-H_A2A: hard-coded A2A baseline
*-MCP: MCP runs

Typical contents of each subfolder:

reference_trajectory.yaml: reference (ground-truth) trajectory/config
evaluate_trajectory*.py: trajectory-metric evaluation (writes CSV)
evaluate_success.py: success-rate computation (writes JSON/CSV)
analyze_retry_patterns.py: retry/error pattern analysis (writes JSON/CSV)

Radar charts

generate_radar_charts.py: generates radar charts from evaluation outputs.

Usage

Run scripts from within the corresponding subfolder, e.g.:

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

Then (from RQ1/):

python3 generate_radar_charts.py