Metraipolis › archive › dc_c3d7cb1cdeca
When Models Know but Can't Linearly Say: Relational Encoding Limits Linear Vulnerability Classification in Code LLMs — arXiv:2610.27257
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.27257
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-21T07:07:37.705Z
- SHA-256 of the archived text
a5781338c13afa271a91ae8d48f6235d554922e461b73b420815621a916eedb6- Stable URL of this page
https://metraipolis.com/documento/dc_c3d7cb1cdeca
How to cite
When Models Know but Can't Linearly Say: Relational Encoding Limits Linear Vulnerability Classification in Code LLMs — arXiv:2610.27257 — arXiv (en), https://arxiv.org/abs/2610.27257 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c3d7cb1cdeca (Metraipolis archive dc_c3d7cb1cdeca, sha256 a5781338c13afa271a91ae8d48f6235d554922e461b73b420815621a916eedb6)
BibTeX
@misc{dcc3d7cb1cdeca,
title = {When Models Know but Can't Linearly Say: Relational Encoding Limits Linear Vulnerability Classification in Code LLMs — arXiv:2610.27257},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.27257},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c3d7cb1cdeca; sha256 a5781338c13afa271a91ae8d48f6235d554922e461b73b420815621a916eedb6},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.27257. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Large language models can detect vulnerable code, yet the form of vulnerability encoding in their representations remains unclear. We analyze 8 code models spanning 6.7B-15B parameters on C programs using three datasets (DELTASECOMMITS, SVEN, and PRECISE-BUGS), comparing two diagnostic tasks on the same representations: identifying what patterns the code instantiates, and determining whether it is vulnerable. Models encode the first with high fidelity. Under the studied paired C-code datasets and linear-probing setup, vulnerability information is more accessible as a paired relational signal-whether code A is riskier than code B-than as a globally separable categorical label. This structural difference has implications for classification: the model reliably ranks which code is riskier but standard classifiers on isolated samples achieve limited accuracy because vulnerable and secure samples form overlapping distributions in representation space. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.27257 — 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 2026-10-21T07:05:43.541Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).