Shoham Sabach

Displacement-Resistant Extensions of DPO with Nonconvex $f$-Divergences

Alignment from Ranking and Rating Information

FisherSFT: Data-Efficient Supervised Fine-Tuning of Language Models Using Information Gain

A Proximal Operator for Inducing 2:4-Sparsity

Comparing Few to Rank Many: Active Human Preference Learning Using Randomized Frank-Wolfe Method

PROXSPARSE: REGULARIZED LEARNING OF SEMI-STRUCTURED SPARSITY MASKS FOR PRETRAINED LLMS