Supplementary Experimental Results for Oracle-Free Metamorphic Test Prioritization Using Bayesian Signal Fusion and Budget-Aware Allocation

Description

This dataset contains supplementary aggregated experimental results, statistical-test outputs, runtime summaries, and Grad-CAM case-evidence summaries supporting the manuscript entitled “Oracle-Free Metamorphic Test Prioritization for Deep Learning Systems Using Bayesian Signal Fusion and Budget-Aware Allocation”. The files support the reported tables and figures, including early Top-N fault exposure, MR entropy, ablation analysis, statistical comparison, and runtime characterization.

Steps to reproduce

The uploaded package contains the aggregated experimental result files used to support the manuscript tables and figures. To reproduce the reported summaries, first read README.txt to identify the spreadsheet associated with each table or figure. The main result file is all_results_multi_seed_fullsuite.xlsx, which contains Top-N selection outputs across datasets, models, seeds, methods, ratios, and budgets. The files aggregated_overall_mean_ci.xlsx, aggregated_average_rank.xlsx, and friedman_nemenyi_results.xlsx provide the mean comparisons, rank summaries, and statistical-test outputs. Runtime results are provided in runtime_cold_cached_summary_updated.xlsx, mode_runtime_summary_fullsuite.xlsx, and gpro_cold_runtime_once_stage_share.xlsx. Grad-CAM case-evidence summaries are provided in gradcam_comparison_summary_fullsuite.xlsx and gradcam_comparison_summary_fullsuite_budgets.xlsx. These files can be inspected directly to verify the manuscript-level tables and figures.

Categories

Funders

• Universiti Teknologi Malaysia

  Grant ID: Q.J130000.3828.23H87

Licence