AILGJul 2

Purified OPSD: On-Policy Self-Distillation Without Losing How to Think

arXiv:2607.0223424.1
Predicted impact top 7% in AI · last 90 daysOriginality Highly original
AI Analysis

For researchers training large language models on reasoning tasks, this work addresses a critical failure mode of OPSD on long-CoT models, offering a principled fix that improves performance without destabilizing reasoning.

On-policy self-distillation (OPSD) fails on long chain-of-thought reasoning models due to a reference-induced component that drives rote memorization. The proposed method, Purified OPSD, isolates the transferable supervision signal using a reference-only teacher and pointwise mutual information, achieving consistent improvements (e.g., +2-5% accuracy) across four models and two datasets while preserving reflective reasoning.

On-policy self-distillation (OPSD) has emerged as a promising paradigm for improving LLM reasoning, where a privileged teacher with access to reference solutions provides token-level supervision on the student's own generated trajectories. However, we find that OPSD consistently fails on long chain-of-thought (long-CoT) reasoning models, yielding at best marginal gains while destabilizing the reflective reasoning capability these models depend on. Through a novel decomposition of the teacher's supervision signal, we identify the root cause: the teacher's supervision is dominated by a reference-induced component that drives rote memorization of reference-specific shortcuts, while the question-conditioned, inference-transferable component is ignored or actively opposed. Based on this diagnosis, we propose a two-step solution. First, we construct a reference-only teacher (the same model conditioned on the reference without the question) to isolate the non-transferable component of the supervision signal; the residual after subtracting this component captures the question-conditioned, inference-transferable correction. Second, we use pointwise mutual information (PMI) as the mechanism to transform this residual into a well-formed PMI target distribution that the student can directly distill from, filtering out the reference-induced shortcut. Experiments on four long-CoT models across two datasets demonstrate consistent improvements over both the base model and standard OPSD, while preserving the models' natural epistemic behavior throughout training.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes