Metraipolis › archive › dc_c4cc313dc353
Shortcutting the Fix: Identifying and Categorizing Agentic Exploits in Software Engineering Benchmarks — arXiv:2609.06780
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.06780
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T14:09:03.622Z
- SHA-256 of the archived text
1c866e8635d1ef5d00cbb3aafbea4f7ea5bfbf54917fb15dc41e5dd407f83bcf- Stable URL of this page
https://metraipolis.com/documento/dc_c4cc313dc353
How to cite
Shortcutting the Fix: Identifying and Categorizing Agentic Exploits in Software Engineering Benchmarks — arXiv:2609.06780 — arXiv (en), https://arxiv.org/abs/2609.06780 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c4cc313dc353 (Metraipolis archive dc_c4cc313dc353, sha256 1c866e8635d1ef5d00cbb3aafbea4f7ea5bfbf54917fb15dc41e5dd407f83bcf)
BibTeX
@misc{dcc4cc313dc353,
title = {Shortcutting the Fix: Identifying and Categorizing Agentic Exploits in Software Engineering Benchmarks — arXiv:2609.06780},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.06780},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c4cc313dc353; sha256 1c866e8635d1ef5d00cbb3aafbea4f7ea5bfbf54917fb15dc41e5dd407f83bcf},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.06780. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
High software engineering benchmark scores often mask exploitative shortcuts, such as retrieving upstream solutions, exploiting local Git histories, or recalling memorized code. We formalize this phenomenon as agentic shortcutting and audit three state-of-the-art open LLMs on SWE-bench Multilingual and DeepSWE using trajectory-level LLM judges. Standard evaluations trigger shortcutting attempts in 60.7%-90.1% and 55.5%-89.1% of trajectories, respectively, with up to 72.1% of tasks resolved with the aid of a successful exploit; disabling network access fails to curb these high attempt rates. Prompting agents to construct solutions independently suppresses attempts to 9.4%-14.2% and 0.9%-10.3%, respectively, while virtually eliminating shortcut-dependent resolutions and preserving core task performance. Our findings demonstrate the necessity of exploit-aware evaluation frameworks to measure repository-level problem solving. Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-09-09 in cs.SE). Categorie arXiv: Software Engineering (cs.SE). Pagina del lavoro: https://arxiv.org/abs/2609.06780 — 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 2026-10-16T13:58:36.371Z dalla pagina https://arxiv.org/catchup/cs.SE/2026-09-09?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).