AI-SOVEREIGN
Home / AI Research / On-Policy Morphing into Hybrid Attention Models
ONAI Research

On-Policy Morphing into Hybrid Attention Models

by ByteDance-Seed · Disen Lan, Xuanda Wang, Jianbin Zheng, Huafeng Kuang, Wei Tang, Yu Cheng

Hybrid attention models combine full and linear attention to achieve a better quality--efficiency trade-off. Converting pretrained Transformers into such models requires identifying which layers should retain full attention under a prescribed budget. We identify two off-policy mismatches in FlashMorph's layer selection: (i) continuous branch interpolation and teacher-forced hidden states differ from discrete hybrid computation; and (ii) frozen model weights fail to account for parameter adaptation during subsequent recovery. We introduce FlashMorph2, an on-policy layer-selection method that (I) evaluates sampled discrete hybrid architectures end to end using their own hidden states, and…

Upvotes13
Comments2
Published17 Nov 2026

More from ByteDance-Seed