arXiv cs.CL· Zhi Xu, Yinqi Wu, Feng Zhang, Xiaotong Zhang, Hong Yu, Han Liu·· 9 小时前AI 评分31
B-OPSD:用自生成问题简报实现同策略自蒸馏
B-OPSD: On-Policy Self-Distillation with Self-Generated Problem Briefs
AI 导读
B-OPSD 提出用模型自身生成的问题简报替代真实答案作为教师模型的特权上下文,训练全程无需 ground-truth 解答。该方法在 AIME 2024、AIME 2025 和 HMMT 2025 上,三个 Qwen3 模型尺寸的平均准确率均高于 OPSD。代码已开源。
正文 · 原文
Abstract:On-Policy Self-Distillation (OPSD) has recently attracted interest because it provides dense teacher guidance without a separate, often larger teacher model. However, OPSD conditions its teacher on ground-truth solutions, which limits its use when such solutions are unavailable. We propose On-Policy Self-Distillation with Self-Generated Problem Briefs (B-OPSD). Using a prompt template, the initial model or a saved student checkpoint generates a problem brief from the question alone. The prompt asks for relevant principles, applicability conditions, and likely pitfalls, but not a final answer or a complete solution. A frozen teacher uses the brief as privileged context to guide the current student's on-policy rollouts through token-level distribution matching. B-OPSD uses no ground-truth solutions during training. Results on AIME 2024, AIME 2025, and HMMT 2025 show that both B-OPSD variants achieve higher average accuracy than OPSD at each of three Qwen3 model sizes. Code is available at this https URL
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.19273 [cs.CL] |
| (or arXiv:2611.19273v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.19273 arXiv-issued DOI via DataCite |
Submission history
From: Zhi Xu [view email]
[v1]
Sat, 14 Nov 2026 09:05:36 UTC (1,356 KB)
[v2]
Tue, 17 Nov 2026 07:55:25 UTC (1,356 KB)
来源:arXiv cs.CL · arxiv.org