CLARITY Replication Package
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Reproducibility artifact for the paper "clarity: From Formal System Designs to Verified RL Controllers." This artifact contains the complete clarity pipeline, which ingests SysML v2 system designs and automatically produces: (1) nuXmv formal models with verified safety properties, (2) shielded reinforcement learning controllers, and (3) runtime safety monitors. All results reported in the paper can be reproduced with a single command. Contents: Three SysML v2 benchmark models (thermostat, cruise control, chemical mixing) SysML-to-SMV extraction and Python simulation twin generator RL training pipeline (behavioral cloning + PPO with specification-derived shield) Runtime monitor evaluation framework nuXmv binary (Linux x86_64, v2.1.0) Pre-computed results: trained checkpoints, verification outputs (IC3 + BMC), and all metrics CSVs Reproducing results: bash run_pipeline.sh --force-full Runs the full pipeline (extract, verify, train, evaluate) in approximately 2 hours. Pre-computed results are included for inspection without re-running. Requirements: Python 3.10+, PyTorch, NumPy, ANTLR4 runtime. See requirements.txt.
Fields of Study & Subjects
Formal Methods in Verification
Safety Systems Engineering in Autonomy
Model-Driven Software Engineering Techniques
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 27 March 2026
- Language
- EN