Metraipolis › archive › dc_02ca107834e1

PersonaManifold: Revealing and Exploiting Curved Geometry in LLM Persona Representations — arXiv:2609.34571

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.34571
License (as archived)
CC0-1.0
Archived
2026-10-07T20:57:55.512Z
SHA-256 of the archived text
0aad74aae98ee2cabe3d3eb6cc9ec69dd209ff00957a482416556682e5120465
Stable URL of this page
https://metraipolis.com/documento/dc_02ca107834e1

How to cite

PersonaManifold: Revealing and Exploiting Curved Geometry in LLM Persona Representations — arXiv:2609.34571 — arXiv (en), https://arxiv.org/abs/2609.34571 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_02ca107834e1 (Metraipolis archive dc_02ca107834e1, sha256 0aad74aae98ee2cabe3d3eb6cc9ec69dd209ff00957a482416556682e5120465)
BibTeX
@misc{dc02ca107834e1,
  title = {PersonaManifold: Revealing and Exploiting Curved Geometry in LLM Persona Representations — arXiv:2609.34571},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.34571},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_02ca107834e1; sha256 0aad74aae98ee2cabe3d3eb6cc9ec69dd209ff00957a482416556682e5120465},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34571. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Controlling persona in large language models (LLMs) at inference time is important for role-playing, personalized dialogue, and social simulation. Recent methods extract persona vectors from the model's activation space and apply Euclidean operations---addition, scaling, and linear interpolation---under the linear representation hypothesis. However, these methods themselves report systematic failures: non-orthogonal trait dimensions, asymmetric ceiling and resistance effects, and significant deviations in multi-trait composition, suggesting that the linear isotropic assumption does not hold. We propose PersonaManifold, a framework that models persona representations as points on a curved, low-dimensional Riemannian submanifold in activation space. We estimate the manifold's intrinsic geometry---local metric tensors, geodesic distances, and Ollivier-Ricci curvature---and introduce geodesic steering, which interpolates between personas along manifold geodesics rather than Euclidean straight lines. We also propose the Behavioral Similarity Triplet (BST) benchmark, which automatically generates situational questions grounded in six established psychological constructs and defines persona similarity through behavioral responses rather than self-report questionnaires. Experiments on three open-source LLMs show that persona activations form a manifold with heterogeneous curvature, geodesic distance predicts behavioral similarity more accurately than Euclidean alternatives with independent contributions from anisotropy and curvature, and geodesic steering produces more coherent intermediate personas on both our BST benchmark and external evaluations, with the advantage concentrated in high-deviation regions where the manifold deviates most from flatness.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.34571 — 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 2026-10-07T20:29:01.591Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).