Skip to content
AI News HubLIVE
Original source2 min read

Subliminal Learning in Vision-Language Models

Summary

A study accepted to WACV 2027 finds that vision-language models can transfer hidden preferences—such as a bias toward a specific animal or facial attribute—from a teacher model to a student model through an unrelated visual task involving geometric shapes and generated numbers. The transfer reaches up to about 10 percentage points in the animal preference task and also shifts how students explain their choices.

SourcearXiv Computer VisionAuthor: Arman Hatami, Romina Aalishah, Ilya E. Monosov
Subliminal Learning in Vision-Language Models
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 4 Nov 2026]

Title:Subliminal Learning in Vision-Language Models

View a PDF of the paper titled Subliminal Learning in Vision-Language Models, by Arman Hatami and 2 other authors

View PDF HTML (experimental)

Abstract:Subliminal learning is a phenomenon in large language models (LLMs) where a teacher passes a hidden preference to a student through training data unrelated to that preference. We study whether subliminal learning also occurs in vision-language models (VLMs). We finetune a teacher VLM to prefer a specific animal or facial attribute and use a control teacher from the same base model. Both teachers generate numbers for an unrelated visual task involving geometric shapes. Two students initialized from the same base VLM are finetuned on the same images, using numbers generated by either the biased or control teacher. Across VLMs ranging from 2B to 7B parameters, we find that both animal and facial preferences can transfer through this unrelated task. Students trained on biased-teacher outputs select the target more often than control students, with a difference of up to $\sim$10 percentage points in the animal preference task. Transfer can become stronger with more training data and varies with the target preference, model, and how the numbers are generated. Beyond preferences, we show that the biased student's explanations use more of the same words and phrases as the biased teacher's explanations than those of the control student, even when all three models select the same image. These results matter for training VLMs and show that training on unrelated data may change both a VLM's image preferences and how it explains its choices, and that transfer can become stronger with more training samples.

Comments: Accepted to WACV 2027

Subjects:

Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Image and Video Processing (eess.IV)

Cite as: arXiv:2611.08538 [cs.CV]

(or arXiv:2611.08538v1 [cs.CV] for this version)

https://doi.org/10.48550/arXiv.2611.08538

arXiv-issued DOI via DataCite

Submission history

From: Arman Hatami [view email] [v1] Wed, 4 Nov 2026 23:36:01 UTC (19,736 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Subliminal Learning in Vision-Language Models, by Arman Hatami and 2 other authors

View PDF

HTML (experimental)

TeX Source

view license

Additional Features

Audio Summary

Current browse context:

cs.CV

new | recent | 2026-11

Change to browse by:

cs cs.AI cs.CL cs.LG eess eess.IV

References & Citations

SciX

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • Hidden preferences in a teacher VLM can transfer to a student VLM through training data from an unrelated visual task
  • Preference transfer was observed across 2B–7B parameter VLMs, with up to ~10 percentage point differences in animal preference
  • Transfer can strengthen with more training data and varies by target preference, model, and number generation method
  • Biased students' explanations share more wording with the biased teacher than control students do, even when all three select the same image

Highlights and analysis are generated automatically and may contain errors. Check the original source.