Metraipolis › archive › dc_f995ae1916bf

Safe Context Switching for Agents in the Wild: Mitigating Subspace Interference via Orthogonal Adaptation — arXiv:2610.05219

Source
arXiv
Original
https://arxiv.org/abs/2610.05219
License (as archived)
CC0-1.0
Archived
2026-10-06T05:25:03.777Z
SHA-256 of the archived text
cfbf4ae9b224b1d7ecd5827206e0aea77778e4d76220a729abfeb27533f62bfe
Stable URL of this page
https://metraipolis.com/documento/dc_f995ae1916bf

How to cite

Safe Context Switching for Agents in the Wild: Mitigating Subspace Interference via Orthogonal Adaptation — arXiv:2610.05219 — arXiv, https://arxiv.org/abs/2610.05219 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f995ae1916bf (Metraipolis archive dc_f995ae1916bf, sha256 cfbf4ae9b224b1d7ecd5827206e0aea77778e4d76220a729abfeb27533f62bfe)
BibTeX
@misc{dcf995ae1916bf,
  title = {Safe Context Switching for Agents in the Wild: Mitigating Subspace Interference via Orthogonal Adaptation — arXiv:2610.05219},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.05219},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f995ae1916bf; sha256 cfbf4ae9b224b1d7ecd5827206e0aea77778e4d76220a729abfeb27533f62bfe},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.05219. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Most Large Language Models exhibit a fundamental tension between two sequential tasks, such as logical reasoning and safety alignment. The high-variance internal states required for sophisticated Chain-of-Thought (CoT) deduction can geometrically interfere with latent representations encoding safety constraints. We identify this phenomenon as Sequential Subspace Interference, showing that standard fine-tuning on logical tasks such as multi-step mathematics and code generation can result in a 23.3% interference penalty on alignment benchmarks, substantially weakening the model's safety priors. This Reasoning Drift is not adequately captured by current adaptation methods because gradients for logical tasks are rarely orthogonal to safety objectives. To address this issue, we propose AURA (Adaptive Unique Residual Allocation), a spectral regularization framework that enforces Spectral Independence between reasoning and safety. By explicitly estimating the null space of the alignment manifold and constraining reasoning updates to its orthogonal complement, AURA enables models to improve logical reasoning without compromising safety. Empirically, AURA recovers 23.0% of the lost performance while preserving greater than 0.98 cosine fidelity to the safe state, demonstrating that reasoning and alignment can be effectively decoupled through geometric regularization.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.05219 — 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 2026-10-06T05:24:05.564Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).