This repository contains an anonymized, review-stage subset of the source code, data, generated rules, and evaluation outputs for Spec2Checker. Author and institution information is intentionally omitted for double-blind review.
| Path | Contents |
|---|---|
Adapter/ |
Adapters for CodeQL, PMD, and Semgrep rule generation |
LLM_to_cases/ |
LLM-based test-case generation scripts |
XML_rules/ |
Generated XML rule artifacts |
data/ |
The currently released evaluation data subset |
baseline/ |
Baseline rule artifacts |
CodeQL_result/ |
CodeQL generation outputs |
PMD_result/ |
PMD generation outputs |
Semgrep_result/ |
Semgrep generation outputs |
validate_tool_result/ |
Syntax-validation outputs |
cost_analysis/ |
Cost-analysis data and figures |
data_summary.md |
Summary of the released data and results |
This is a partial preview rather than the complete archival artifact. Additional pipeline components, datasets, and experiment outputs are being organized and will be included in the final public release upon acceptance.
- Generated rule files are preserved as raw experimental outputs. Some files are intentionally syntactically invalid because rule-generation validity is one of the evaluated outcomes.
- Credentials are not included. Scripts that call an LLM read configuration from
environment variables; see
.env.example. - Some scripts depend on pipeline components that are not part of this partial preview and therefore are not yet runnable as a complete end-to-end system.
The Python packages used by the released scripts are listed in
requirements.txt. Install them with:
python -m pip install -r requirements.txtAn anonymized snapshot of our source code and data is available in this repository. The final archival version will be made publicly available upon acceptance.