When self-distillation with a privileged-information teacher is used alone on hard tasks, the loss decreases but accuracy does not, because the teacher transfers one solution rather than correctness.
On-policy distillation of language models: Learning from self- generated mistakes
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Privileged, but Biased: How PI-Conditioned Teachers Break Self-Distillation
When self-distillation with a privileged-information teacher is used alone on hard tasks, the loss decreases but accuracy does not, because the teacher transfers one solution rather than correctness.