Ariadne operates on a source base model and its fine-tuned counterpart, and a target base model of different depth or width.
A training-free method transfers fine-tuned behavior across heterogeneous Transformers
Ariadne uses activation changes rather than parameter transport to construct native task vectors for target models of different depth and width.
Academic
Filippo Rinaldi · Aniello Panariello · Giacomo Salici · Simone Magistri · Angelo Porrello · Andrew D. Bagdanov · +1 more
University of Modena and Reggio Emilia · University of Florence
Research Digest··3 min read
The authors propose Ariadne, a training-free approach that transfers task-specific behavior across Transformer architectures by focusing on the activation changes induced by fine-tuning rather than moving parameters directly.
The Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.