Seed-Induced Uniqueness in Transformer Models: Subspace Alignment Governs Subliminal Transfer
We analyze subliminal transfer in Transformer models, where a teacher embeds hidden traits that can be linearly decoded by a student without degrading main-task performance. Prior work often attributes transferability to global representational simila
Authors
- Laxima Niure Kandel (ORCID: https://orcid.org/0000-0002-3933-1187)
- Berker Peköz (ORCID: https://orcid.org/0000-0002-7572-3663)
- Mustafa İlhan Akbaş (ORCID: https://orcid.org/0000-0002-5450-3522)
- Ayse S Okatan
Publication Details
- Published
- 2025-10-14
- DOI
- https://doi.org/10.36227/techrxiv.176045931.11179043/v1
- Primary Topic
- Islanding Detection in Power Systems
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Science Foundation