Metraipolis › archive › dc_74e3eaef39dc

LLMs as Linguistic Chameleons: Decoupling Semantics and Structure for Privacy-Preserving Communication — arXiv:2609.23193

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.23193
License (as archived)
CC0-1.0
Archived
2026-10-09T06:50:32.747Z
SHA-256 of the archived text
44794c090acfc10754b12e87c54f626ce85c1fe9c176ba9f3c6eb82d4cb616a2
Stable URL of this page
https://metraipolis.com/documento/dc_74e3eaef39dc

How to cite

LLMs as Linguistic Chameleons: Decoupling Semantics and Structure for Privacy-Preserving Communication — arXiv:2609.23193 — arXiv (en), https://arxiv.org/abs/2609.23193 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_74e3eaef39dc (Metraipolis archive dc_74e3eaef39dc, sha256 44794c090acfc10754b12e87c54f626ce85c1fe9c176ba9f3c6eb82d4cb616a2)
BibTeX
@misc{dc74e3eaef39dc,
  title = {LLMs as Linguistic Chameleons: Decoupling Semantics and Structure for Privacy-Preserving Communication — arXiv:2609.23193},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.23193},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_74e3eaef39dc; sha256 44794c090acfc10754b12e87c54f626ce85c1fe9c176ba9f3c6eb82d4cb616a2},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.23193. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

As Large Language Model (LLM) APIs become increasingly integrated into privacy-sensitive workflows, ensuring inference-time privacy without compromising task utility remains a major challenge. Existing approaches preserve most of the original semantic content to maintain downstream performance, but this also leaves exploitable cues for reconstructing the original text. This work investigates semantic decoupling, which replaces original semantics with alternative content while preserving the structure needed for LLM reasoning. Based on this idea, we propose CROSS-MAP, a bidirectional framework that maps private inputs into a different semantic domain before inference and recovers the corresponding outputs afterward. Local models are trained with multi-objective optimization to maximize semantic divergence in the mapping stage while minimizing semantic inconsistency in the recovery stage. Experiments show that CROSS-MAP reduces reconstruction success across multiple attack settings while outperforming existing baselines in utility.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-22 in cs.CL).
Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.23193 — 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 2026-10-09T06:48:45.116Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).