Dataset for the paper "BRIDGE: Agentic Support from Feature Change to Evidence-Grounded ADRs"

This dataset is associated with the paper titled “BRIDGE: Agentic Support from Feature Change to Evidence-Grounded ADRs”. It is stored in a Microsoft Excel file that consists of several worksheets. Below is a brief description of each worksheet. (1) The ‘Overview’ worksheet provides an overview of the study design, the four research questions, Large Language Model (LLM) configurations, and the location of each published artifact. (2) The ‘Sheet Guide’ worksheet provides a one-line description of every worksheet included in the workbook. (3) The ‘Master_Raw_Data’ worksheet contains all RQ1–RQ5 metrics for the 200 experimental runs, joined using the run_id identifier across 87 columns. The data can be filtered by system, feature, and model. (4) The ‘RQ1_Raw_Data’ worksheet contains system-level architectural-evidence metrics that provide the source data for Table II. (5) The ‘RQ1_Final_Analysis’ worksheet presents Table II, which reports the architectural evidence recovered for each system, followed by the supporting analysis tables for RQ1-1 and RQ1-2. (6) The ‘RQ2_Raw_Data’ worksheet contains long-form data on direct and transitive impact occurrences for each run and architectural element type. It also includes a flag identifying the representative run. (7) The ‘RQ2_Final_Analysis’ worksheet presents Table III, which reports the distribution of impacts by architectural element type, followed by per-feature results for the representative model. (8) The ‘RQ3_Raw_Data’ worksheet contains the structural-integrity and traceability checks for all 200 experimental runs, providing the source data for Table IV. (9) The ‘RQ3_Final_Analysis’ worksheet presents Table IV, including (a) run-level satisfaction results by model and (b) artifact-level coverage results. (10) The ‘RQ4_Raw_Data’ worksheet contains per-run evaluation findings, refinement actions, and before-and-after metrics with their corresponding differences. It provides the source data for Figure 2. (11) The ‘RQ4_Final_Analysis’ worksheet presents Figure 2, which reports alternatives with missing trace links by model, followed by detailed per-feature and cross-model results.