The Master Key Hypothesis: Unlocking Cross-Model Capability Transfer via Linear Subspace Alignment
arXiv:2604.06377v1 Announce Type: new Abstract: We investigate whether post-trained capabilities can be transferred across models without retraining, with a focus on transfer across different model …
Rishab Balasubramanian, Pin-Jie Lin, Rituraj Sharma, Anjie Fang, Fardin Abdi, Viktor Rozgic, Zheng Du, Mohit Bansal, Tu Vu
11 views