2 papers
eess.SP2025
Seed-Induced Uniqueness in Transformer Models: Subspace Alignment Governs Subliminal Transfer
AyÅe Selin Okatan, Mustafa Ä°lhan AkbaÅ, Laxima Niure Kandel +1
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…
cs.CR2025
Keys in the Weights: Transformer Authentication Using Model-Bound Latent Representations
AyÅe S. Okatan, Mustafa Ä°lhan AkbaÅ, Laxima Niure Kandel +1
We introduce Model-Bound Latent Exchange (MoBLE), a decoder-binding property in Transformer autoencoders formalized as Zero-Shot Decoder Non-Transferability (ZSDN). In identity tas…