arXiv · cs.AI · Suwan Wu, Yumeng Lin, Pengcheng Yuan, Xiaolong Jiang2026-11-02 13:00 · 10 小时前模型评分28

统一逐 token 门控家族：FKL/RKL 混合与多通道、偏置系数用于在线策略蒸馏

A Unified Per-Token Gating Family for On-Policy Distillation: FKL/RKL Mixing with Multi-Channel and Bias Coefficients

导读

研究提出四系数逐 token 门控参数化 λ_t = σ(a·h_t + b·u(x) + c + d·gap_t)，把 EOPD 与 ToDi 的方向对齐代理纳入同一坐标系，并新增多通道组合与显式偏置。

分类置信度

中 · 待人工复核

来源级别：媒体与个人

事件类型：研究发布

证据摘录

“We introduce a four-coefficient parameterization lambda_t = sigma(a * h_t + b * u(x) + c + d * gap_t) in which direction-aligned proxies of EOPD and ToDi appear as one-dimensional (1D) restrictions, and which adds multi-channel composition and an explicit bias as further degrees of freedom.”

来源：arXiv · cs.AI · arxiv.org