Hindsight Compacts but Does Not Repair: Rethinking On-Policy Self-Distillation in Reasoning Models

arXiv:2605.06188v2 Announce Type: replace-cross Abstract: On-Policy Self-Distillation (OPSD) has emerged as a promising post-training method: requiring only privileged context, it enables token-level credit assignment from a self-teacher with hindsight, yielding higher accuracy with shorter responses. However, in thinking-enabled mathematical reasoning, where solving a problem takes long reasoning traces that explore, reflect, and backtrack, its length reductions persist but its accuracy gains largely do not. Does OPSD's hindsight signal, then, still repair failed trajectories, or does it only compact viable ones? We hypothesize compaction: in long traces, hindsight reveals which steps were unnecessary more readily than which steps would have repaired the solution. To test this, we apply OPSD separately to correct and incorrect rollouts. Training only on correct rollouts shortens responses by 18-29% while largely preserving accuracy, whereas training only on incorrect rollouts degrades it. The accuracy gap holds across three models, six benchmarks, and three seeds, and under a same-prompt control for difficulty. Neither branch raises the pass@k ceiling, and both suppress exploration and reflection markers, which viable traces can spare but failed ones need. Changing the divergence, enriching or reinjecting the privileged context, and training longer only move OPSD along the same accuracy-length tradeoff. Hindsight compacts reasoning the model can already produce but does not repair reasoning it cannot.

Abstract:On-Policy Self-Distillation (OPSD) has emerged as a promising post-training method: requiring only privileged context, it enables token-level credit assignment from a self-teacher with hindsight, yielding higher accuracy with shorter responses. However, in thinking-enabled mathematical reasoning, where solving a problem takes long reasoning traces that explore, reflect, and backtrack, its length reductions persist but its accuracy gains largely do not. Does OPSD's hindsight signal, then, still repair failed trajectories, or does it only compact viable ones? We hypothesize compaction: in long traces, hindsight reveals which steps were unnecessary more readily than which steps would have repaired the solution. To test this, we apply OPSD separately to correct and incorrect rollouts. Training only on correct rollouts shortens responses by 18-29% while largely preserving accuracy, whereas training only on incorrect rollouts degrades it. The accuracy gap holds across three models, six benchmarks, and three seeds, and under a same-prompt control for difficulty. Neither branch raises the pass@k ceiling, and both suppress exploration and reflection markers, which viable traces can spare but failed ones need. Changing the divergence, enriching or reinjecting the privileged context, and training longer only move OPSD along the same accuracy-length tradeoff. Hindsight compacts reasoning the model can already produce but does not repair reasoning it cannot.

Artificial Intelligence (cs.AI); Computation and Language (cs.CL)

arXiv:2605.06188 [cs.AI]

(or arXiv:2605.06188v2 [cs.AI] for this version)

https://doi.org/10.48550/arXiv.2605.06188 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

aiqbench analysis

Hindsight Compacts but Does Not Repair: Rethinking On-Policy Self-Distillation in Reasoning Models

Related coverage