${\mathrm{E}}^2$-$\mathrm{OPSD}$: Taming Entropy Overshoot in On-Policy Self-Distillation

On-policy self-distillation (OPSD) provides dense token-level supervision without a second model: one network acts as teacher with the reference solution and as student with only the problem. We identify a specific failure mode of this recipe. During training, student token entropy rises past the teacher’s and remains elevated, a pattern we call entropy overshoot. We trace it to both sides of distillation. The reference-conditioned teacher is confident along its answer-directed reasoning path, but this confidence transfers poorly to student-generated prefixes, making its supervision overly tied to answer-specific cues rather than reusable reasoning patterns; meanwhile, the forward KL used by OPSD continually diffuses the student’s predictive distribution without pulling it back. We introduce ${\mathrm{E}}^2$-$\mathrm{OPSD}$ to address both causes. Exemplar-guided teaching replaces the current answer with a retrieved solved neighboring problem, providing transferable reasoning guidance without revealing the destination and better matching student- reachable states. Entropy-aware distillation uses the student–teacher entropy gap to determine the direction and strength of each token’s correction. ${\mathrm{E}}^2$-$\mathrm{OPSD}$ improves math reasoning by up to 4.3 points in mean@16 over OPSD, while out- of-domain evaluations show gains over the corresponding base models of up to 4.9 points in mean@16 and 5.5 points in pass@8. Despite these gains, ${\mathrm{E}}^2$-$\mathrm{OPSD}$ remains simple, requiring no additional forward passes or networks during distillation.