Kryptovalutaticker:
technology från Arxiv cs.ai

Quantifying Subliminal Behavioral Transfer Ratios in Language Model Distillation

Uwe Konig, Hamza Kazmi, Ruizhe Li, Maheep Chaudhary
Thursday at 04:00
4 Visningar
0 Kommentarer

arXiv:2606.11270v1 Announce Type: cross Abstract: Distillation of a language model intended to transfer benign behavior to a student model may also transfer undesirable characteristics, if they are present in the teacher model, a phenomenon known as subliminal learning. While qualitative evidence supports the existence of this effect, its...

Läs hela artikeln hos källan.

Var detta hjälpsamt?
Dela:

Kommentarer (0)

Vänligen logga in för att publicera en kommentar

Inga kommentarer ännu. Bli först med att kommentera!