SeerAttention-R: Sparse Attention Adaptation for Long Reasoning
Gao, Yizhao, Guo, Shuming, Cao, Shijie, Xia, Yuqing, Cheng, Yu, Wang, Lei, Ma, Lingxiao, Sun, Yutao, Ye, Tianzhu, Dong, Li, So, Hayden Kwok-Hay, Hua, Yu, Cao, Ting, Yang, Fan, Yang, Mao
–arXiv.org Artificial Intelligence
We introduce SeerAttention-R, a sparse attention framework specifically tailored for the long decoding of reasoning models. Extended from SeerAttention, SeerAttention-R retains the design of learning attention sparsity through a self-distilled gating mechanism, while removing query pooling to accommodate auto-regressive decoding. With a lightweight plug-in gating, SeerAttention-R is flexible and can be easily integrated into existing pretrained model without modifying the original parameters. We demonstrate that SeerAttention-R, trained on just 0.4B tokens, maintains near-lossless reasoning accuracy with 4K token budget in AIME benchmark under large sparse attention block sizes (64/128). Using TileLang, we develop a highly optimized sparse decoding kernel that achieves near-theoretical speedups of up to 9x over FlashAttention-3 on H100 GPU at 90% sparsity. Code is available at: https://github.com/microsoft/SeerAttention.
arXiv.org Artificial Intelligence
Jun-11-2025