ADEPT: A Unified Framework for Deep Learning Test Adequacy
ADEPT packages multiple deep-learning test adequacy methods behind one workflow.
The paper says existing adequacy metrics are hard to reproduce and compare because they arrive as separate prototypes with different setup, preprocessing, and configuration requirements. ADEPT integrates representative approaches including neuron coverage, surprise adequacy, input distribution coverage, boundary coverage, and mutation scores. It adds a template-based metric interface, YAML configuration, preprocessing-cache reuse, and structured results for research and development use. ArXiv · AI/CL/LG's note
The paper says existing adequacy metrics are hard to reproduce and compare because they arrive as separate prototypes with different setup, preprocessing, and configuration requirements. ADEPT integrates representative approaches including neuron coverage, surprise adequacy, input distribution coverage, boundary coverage, and mutation scores. It adds a template-based metric interface, YAML configuration, preprocessing-cache reuse, and structured results for research and development use. ArXiv · AI/CL/LG's note
score 4