Showing cs.LGShow all
2 papers · 1 filter
cs.LG2026
Does On-Policy Distillation Need to Complete Every Rollout?
Zhengxin Zhang, Maosen Tang, Ning Wang +6
On-policy distillation (OPD) transfers teacher knowledge to a student through feedback on student-generated rollouts, but generating complete rollouts and computing per-token teach…
cs.LG2026
HARP: Efficient Data Selection for Finetuning Large Language Models
Ning Wang, Zhengxin Zhang, Maosen Tang +3
Finetuning data selection requires balancing two competing goals: selecting examples that improve the downstream objective, and doing so without repeatedly finetuning models. Train…