Unverified50% confidenceFactExact time
LEAF-OPD gives the teacher model an additional instruction (privileged context) on forget data to guide the student to learn to refuse, while on retain data the teacher does not use privileged context; the student always runs without privileged context.
1
Sources
50%
Confidence
Long-term
Relevance
11/17/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPrivileged information in self-distillation refers to information that the teacher model can access but that is unavailable to the student at inference time, such as ground-truth answers, a longer reasoning budget, or additional verification signals.73% similarUnverifiedUsing only the student model's mastery level as the admission criterion would exclude useful teacher demonstrations.70% similarUnverifiedSynthetic data distillation prevents the student model from learning the teacher's confidence distribution, resulting in relatively limited generalization ability on ambiguous samples.67% similarUnverifiedThe capability of supervised learning models is bounded by the ceiling of data quality and cannot surpass the level of the teacher algorithm.67% similarUnverifiedOff-policy distillation directly uses trajectories generated by the teacher; the student is exposed to the teacher's distribution, which deviates from its own test-time behavior and can easily cause forgetting or degraded generalization.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1179600API
curl https://kongchang.com/api/v1/knowledge/claims/1179600MCP
get_claim(id=1179600)