Metraipolis › archive › dc_8d55ffc16fc7
APG: Agentic Static-Analysis Infrastructure on LLVM-IR Graph Databases — arXiv:2611.02349
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.02349
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-03T11:27:20.163Z
- SHA-256 of the archived text
52122269ffadcb8f4e9700dcf80ffbe1a9850c6d3f6419e0b463966e089071d1- Stable URL of this page
https://metraipolis.com/documento/dc_8d55ffc16fc7
How to cite
APG: Agentic Static-Analysis Infrastructure on LLVM-IR Graph Databases — arXiv:2611.02349 — arXiv (en), https://arxiv.org/abs/2611.02349 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8d55ffc16fc7 (Metraipolis archive dc_8d55ffc16fc7, sha256 52122269ffadcb8f4e9700dcf80ffbe1a9850c6d3f6419e0b463966e089071d1)
BibTeX
@misc{dc8d55ffc16fc7,
title = {APG: Agentic Static-Analysis Infrastructure on LLVM-IR Graph Databases — arXiv:2611.02349},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.02349},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8d55ffc16fc7; sha256 52122269ffadcb8f4e9700dcf80ffbe1a9850c6d3f6419e0b463966e089071d1},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.02349. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
LLM-based coding agents increasingly rely on external program knowledge, yet access to static-analysis graphs alone does not ensure appropriate operation selection or evidence interpretation. We present APG, an infrastructure for semantics-grounded static program analysis. Built on SVF and LLVM, it decomposes compiler-derived program analyses into atomic capabilities that LLM agents can select, compose, and orchestrate for goal-directed, cross-graph investigation, with query safeguards and evidence traces preserving inspectable results and provenance. APG provides 38 atomic capabilities. On analysis requests arising during benchmark repair runs, 35 achieve first-attempt query-generation success rates above 50%, including 13 at 100%. Evaluation on 812 C/C++ defects shows higher repair success than the evaluated baselines. Compared with Passive Graph Push, APG improves first-iteration repair success by 17.6 and 18.0 percentage points on the two benchmark collections and reduces average LLM input-token consumption by 90.8%. Our tool is available at this https URL and its demonstration video at this https URL Tema degli esploratori: calcolo e infrastrutture (elenco arXiv cs.SE, nuovi lavori del giorno). Categorie arXiv: Software Engineering (cs.SE). Pagina del lavoro: https://arxiv.org/abs/2611.02349 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-11-03T11:02:52.195Z dalla pagina https://arxiv.org/list/cs.SE/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).