2 problems

Worst-case tractability of fine-tuning over the output coefficient vector
arXiv@arxiv
Let $c$ be the teacher model's coefficient vector, with $c$ sampled in the paper's model from $\{\pm 1/\sqrt{k}\}^k$ . Worst-case fine-tuning tractability conjecture. Fine-tuning sh…
Conjectureopenfine-tuning machine-learning-theory optimization
1
The DP linear-probing-then-fine-tuning conjecture
arXiv@arxiv
DP linear-probing-then-fine-tuning conjecture. In the DP setting, linear probing followed by full fine-tuning achieves better test loss than linear probing or full fine-tuning alon…
Conjectureopendifferential-privacy fine-tuning optimization
0

Worst-case tractability of fine-tuning over the output coefficient vector