Metraipolis › archive › dc_2afed46fd147

Fisher-IRG: Fisher-Induced Local Invariant Representation Geometry across Language and Vision Models — arXiv:2609.36458

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.36458
License (as archived): CC0-1.0
Archived: 2026-10-07T16:57:26.343Z
SHA-256 of the archived text: e6977276a9c43ba619a3b06846aa7b91d024ea6e05a1f0e0962c9fb59507aeb1
Stable URL of this page: https://metraipolis.com/documento/dc_2afed46fd147

How to cite

Fisher-IRG: Fisher-Induced Local Invariant Representation Geometry across Language and Vision Models — arXiv:2609.36458 — arXiv (en), https://arxiv.org/abs/2609.36458 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2afed46fd147 (Metraipolis archive dc_2afed46fd147, sha256 e6977276a9c43ba619a3b06846aa7b91d024ea6e05a1f0e0962c9fb59507aeb1)

BibTeX

@misc{dc2afed46fd147,
  title = {Fisher-IRG: Fisher-Induced Local Invariant Representation Geometry across Language and Vision Models — arXiv:2609.36458},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.36458},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2afed46fd147; sha256 e6977276a9c43ba619a3b06846aa7b91d024ea6e05a1f0e0962c9fb59507aeb1},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.36458. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Semantic-preserving transformations can induce substantial motion in learned representations, while small changes may strongly affect model predictions, raising a basic question: what local metric best captures semantically consequential variation? We propose Fisher-induced invariant representation geometry (Fisher-IRG), which measures local representation directions through their predictive sensitivity. Around each representation, we construct semantic-preserving and semantic-changing neighborhoods, aggregate their local Fisher information, and recover invariant directions through a contrastive generalized eigenvalue problem. Controlled displacement analyses first show that comparable Euclidean motion can have substantially different predictive consequences, supporting the need for a predictive geometry. Across language and vision models, Fisher-IRG yields stronger semantic-versus-nuisance predictive selectivity and generally more reproducible subspaces than covariance-based geometry, while recovering systematically distinct local directions. Representation interventions further localize semantic effects to the Fisher-derived subspace, and held-out separation and retrieval show that the recovered geometry generalizes beyond the discovery neighborhoods. These results support Fisher-IRG as a principled framework for characterizing local invariant representation geometry.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.CL).
Categorie arXiv: Machine Learning (cs.LG) ; Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.36458 — 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 2026-10-07T16:27:03.399Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).