Metraipolis › archive › dc_477d3e4c588d

Probing Perceptual Priors of MLLMs via Gibbs Sampling with Interpretable Generative Controls — arXiv:2608.27727

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.27727
License (as archived)
CC0-1.0
Archived
2026-10-17T10:47:10.364Z
SHA-256 of the archived text
b6159e87ad8fc9b2bf376186794f2376cfcdc0ffde94ff7be33f8d9065f0865f
Stable URL of this page
https://metraipolis.com/documento/dc_477d3e4c588d

How to cite

Probing Perceptual Priors of MLLMs via Gibbs Sampling with Interpretable Generative Controls — arXiv:2608.27727 — arXiv (en), https://arxiv.org/abs/2608.27727 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_477d3e4c588d (Metraipolis archive dc_477d3e4c588d, sha256 b6159e87ad8fc9b2bf376186794f2376cfcdc0ffde94ff7be33f8d9065f0865f)
BibTeX
@misc{dc477d3e4c588d,
  title = {Probing Perceptual Priors of MLLMs via Gibbs Sampling with Interpretable Generative Controls — arXiv:2608.27727},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.27727},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_477d3e4c588d; sha256 b6159e87ad8fc9b2bf376186794f2376cfcdc0ffde94ff7be33f8d9065f0865f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.27727. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

A model's behavior on a task is jointly determined by the input it receives and the prior it brings in, i.e. the distribution over stimuli it implicitly expects. Interpretability research has traditionally studied models by holding inputs fixed and examining model responses either mechanistically, probing how internal structure represents inputs, or behaviorally, measuring how variation in inputs leads to variation in outputs. Neither reconstructs the prior distribution itself, since internal structure shows what a model can represent, not what it expects, and any fixed stimulus set leaves most of the possible input space unseen. In particular, such an input space in real-world settings, such as images seen by VLMs, is extremely high-dimensional and diverse. These priors thus remain a poorly understood component of models that nonetheless influence real-world behavior. We propose a method to sample from models' perceptual prior distributions directly, by steering a generative model to produce stimuli along controllable axes and running Gibbs sampling over that space with the model under study as the judge. We apply this to a variety of categories and target variables (such as trustworthiness in faces and cheapness in art images) and recover both canonical biases and surprising novel priors invisible to direct prompting, warranting further investigation of their downstream effects.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-31 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.27727 — 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 2026-10-17T10:33:42.344Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-31?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).