paper-with-me

홈 › Papers

Near-Policy: Accelerating On-Policy Distillation via Asynchronous Generation and Selective Packing

2026-05-07 · Miao Rang, Zhenni Bi, Hang Zhou, Kai Han, Xuechun Wang, An Xiao, Xinghao Chen, Yunhe Wang, Hanting Chen arxiv

Standard knowledge distillation for autoregressive models often suffers from distribution mismatch. While on-policy methods mitigate this by leveraging student-generated outputs, they rely on computationally expensive Reinforcement Learning (RL) frameworks. To improve efficiency, we propose Near-Policy Distillation (NPD), an asynchronous approach that decouples student generation from training. This reformulation enables Supervised Fine-Tuning (SFT) with sequence packing. However, asynchronous updates inevitably introduce policy lag and sample noise, which can cause the behavior to drift from near-policy toward off-policy. To counteract this without sacrificing efficiency, NPD integrates sparse student updates and the $Δ$-IFD filtering mechanism, a heuristic sample selection mechanism that empirically stabilizes the optimization trajectory. By filtering extreme out-of-distribution samples, $Δ$-IFD prevents noise from dominating the gradients, ensuring updates remain within a safe proximal learning zone. Empirically, the NPD framework achieves a 8.1x speedup over on-policy baselines and outperforms SFT by 8.09%. Crucially, by effectively narrowing the exploration space for subsequent RL, our method enables openPangu-Embedded-1B to reach a state-of-the-art score of 68.73%, outperforming the substantially larger Qwen3-1.7B. Codes will be released soon.

📄 PDF Abstract BibTeX arXiv:2605.05940

Code (0)

등록된 구현이 없습니다.

Tasks

Knowledge DistillationReinforcement Learning

Similar Papers 제목 키워드 기반

A-3PO: Accelerating Asynchronous LLM Training with Staleness-aware Proximal Policy Approximation

2025-12-06 · Xiaocan Li, Shiliang Wu, Zheng Shen arxiv

Decoupled PPO has been a successful reinforcement learning (RL) algorithm to deal with the high data staleness under the asynchronous RL setting. Decoupled loss used in decoupled PPO improves coupled-loss style of algori…

Reinforcement Learning

$\boldsymbol{f}$-OPD: Stabilizing Long-Horizon On-Policy Distillation with Freshness-Aware Control

2026-05-18 · Xianwei Chen, Shimin Zhang, Jibin Wu arxiv

Scaling on-policy distillation (OPD) for large language models (LLMs) confronts a fundamental tension: asynchronous execution is necessary for system efficiency, but structurally deviates from the ideal on-policy objecti…

Periodic Asynchrony: An On-Policy Approach for Accelerating LLM Reinforcement Learning

2025-11-24 · Jian Lu arxiv

Since the introduction of the GRPO algorithm, reinforcement learning (RL) has attracted increasing attention for LLM post-training, yet training efficiency remains a critical challenge. In mainstream RL frameworks, infer…

Reinforcement Learning

AsyncOPD: How Stale Can On-Policy Distillation Be?

2026-06-23 · Wonjun Kang, Kevin Galim, Seunghyuk Oh, Minjun Kang 외 arxiv

On-policy distillation (OPD) trains a student on its own rollouts guided by teacher feedback and is becoming increasingly important for large language model (LLM) post-training. Like reinforcement learning (RL), however,…

Reinforcement Learning

Eliciting Weak-to-Strong Generalization with On-Policy Reverse Distillation

2026-09-08 · Youngrok Park, Sangmin Bae, Hojung Jung, Jongwoo Ko 외 hf

Weak-to-strong generalization asks whether stronger models can learn from weaker supervisors and surpass them. This question is particularly important for successive model generations and multi-domain consolidation, wher…