| title | Denser neq Better: Limits of On-Policy Self-Distillation for Continual Post-Training |
|---|---|
| authors | Meng Wang, Haohan Zhao, Wenzhuo Liu, Lu Yang, Geng Liu, Haiyang Guo, Guo-Sen Xie, Gaofeng Meng, Hongbin Liu, Fei Zhu |
| date | 2026-07-02 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2607.01763 |
| upvotes | 6 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Continual post-training enables foundation models to acquire new knowledge while preserving existing capabilities. Recent work suggests that on-policy learning can mitigate forgetting, with on-policy self-distillation emerging as a particularly attractive approach. In this work, we revisit this optimistic view through self-distillation policy optimization (SDPO). Our experiments show that SDPO can accelerate in-domain specialization when teacher signals are stable and well aligned, but it struggles to generalize to out-of-distribution scenarios. In continual post-training, SDPO exhibits stronger forgetting and can even collapse, whereas on-policy reinforcement learning methods such as GRPO adapt more conservatively and better preserve prior capabilities. Further analyses reveal that denser self-distillation induces larger drift in both parameter space and response space, and can amplify high-frequency formatting artifacts through a self-reinforcing teacher--student loop. These findings suggest that on-policy data alone is insufficient for continual learning. Dense self-distillation can accelerate specialization when teacher targets are stable and token-level supervision is reliable, but it should not be treated as a default stabilizer for continual post-training. Our code is available at https://github.com/Moenupa/SDPO-CL. |
cite
@misc{wang2026denser,
title={Denser neq Better: Limits of On-Policy Self-Distillation for Continual Post-Training},
author={Meng Wang and Haohan Zhao and Wenzhuo Liu and Lu Yang and Geng Liu and Haiyang Guo and Guo-Sen Xie and Gaofeng Meng and Hongbin Liu and Fei Zhu},
year={2026},
eprint={2607.01763},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2607.01763}
}