Every Coin Has Two Sides: On the Dual Nature of Generalization in On-Policy Distillation of Large Language Models
On-policy distillation (OPD) transfers teacher capabilities by supervising trajectories sampled from the student's own policy, yet its generalization behavior remains poorly understood, as most studies evaluate OPD on a single domain and on benchmarks close to the training data. We present a controlled study that varies one generalization factor at a time, from in-domain distribution shifts to cross-domain transfer and the multi-teacher setting. We find that OPD transfers a teacher's reasoning behavior rather than its answers to particular problems: training difficulty barely matters, and even
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- FuzzyOverlapping authors or contributors · 62%Kong/kong →
“Shared author/contributor keys: kong”
- LinkedLinked via arxiv author · 85%Zhaoyi Li →
“Every Coin Has Two Sides: On the Dual Nature of Generalization in On-Policy Distillation of Large Language Models”
- LinkedLinked via arxiv author · 85%Deyang Kong →
“Every Coin Has Two Sides: On the Dual Nature of Generalization in On-Policy Distillation of Large Language Models”
- LinkedLinked via arxiv author · 85%Xueyuan Wei →
“Every Coin Has Two Sides: On the Dual Nature of Generalization in On-Policy Distillation of Large Language Models”
- LinkedLinked via arxiv author · 85%Evan Yang →
“Every Coin Has Two Sides: On the Dual Nature of Generalization in On-Policy Distillation of Large Language Models”
- LinkedLinked via arxiv author · 85%Ranran Shen →
“Every Coin Has Two Sides: On the Dual Nature of Generalization in On-Policy Distillation of Large Language Models”
- LinkedLinked via arxiv author · 85%Mahardika Krisna Ihsani →
“Every Coin Has Two Sides: On the Dual Nature of Generalization in On-Policy Distillation of Large Language Models”
- LinkedLinked via arxiv author · 85%Ming Yang →
“Every Coin Has Two Sides: On the Dual Nature of Generalization in On-Policy Distillation of Large Language Models”