Not Every Token Is Worth Distilling: Selective Supervision for Direct-OPD
Selective Direct-OPD keeps the top 10% of high-divergence states and beats dense distillation on math benchmarks.
1 stories, newest first · one card per story
Selective Direct-OPD keeps the top 10% of high-divergence states and beats dense distillation on math benchmarks.
Today 1.6K readers · 2.5K views · 779 reads · all time 189K views