# FiRe-OPD：先过滤，再重加权--重新思考在线策略蒸馏的优化粒度

- 来源：HuggingFace Daily Papers（社区热门论文）
- 发布时间：2026-06-01 08:00
- AIHOT 分数：57
- AIHOT 链接：https://aihot.virxact.com/items/cmpz0edm20523sli3uoo6y5aw
- 原文链接：https://arxiv.org/abs/2606.02684

## AI 摘要

FiRe-OPD（Filter, then Reweight）重新思考在线策略蒸馏的优化粒度，在轨迹和token两个层面联合调整监督信号。先过滤低质量轨迹，再对保留轨迹内的token进行软加权，避免硬选择带来的信息损失并提升优化稳定性。该方法在强到弱、单教师、多教师三种设置下均优于近期token级OPD方法：在AIME 2024上提升6.25分，在Miner上提升18.81分。代码已开源。

## 正文

On-Policy distillation (OPD) in large language models is shifting from full-trace KL supervision toward more selective training paradigms. Recent OPD methods increasingly focus on selecting which trajectories to learn from, which tokens are most informative, and which supervision signals are most reliable. Motivated by this trend, we rethink optimization granularity of OPD and propose \fireicon\ FiRe-OPD (Filter, then Reweight), which jointly adjusts supervision signals at both trajectory and token levels. In details, FiRe-OPD first filters trajectories to remove low-quality rollout samples, and then applies soft reweighting within the retained trajectories to emphasize informative tokens. Compared with hard token selection, FiRe-OPD leverages a soft-weighting mechanism to effectively mitigate information loss and enhance optimization stability, thereby achieving finer-grained OPD optimization. We validate the effectiveness of FiRe-OPD across strong-to-weak, single-teacher, and multi-teacher settings, and demonstrate its superiority over recent token-level OPD methods ( (e.g., +6.25 on AIME 2024 in strong-to-weak, +18.81 on Miner in multi-teacher). Our code is available at https://github.com/YuYingLi0/FiRe-OPD.