Microsoft Paper: Sliding Window Attention Beats Linear Attention for Inference Memory

rohanpaul_ai · x · 2026-09-02

A new Microsoft paper finds that for reducing inference memory, training-free Sliding Window Attention (SWA) outperforms most retrofitted linear-attention methods.

Methodology: Keep only a small recent window plus the first 4 “sink” tokens that models rely on.

Key Findings:

Conclusion: While full attention still wins on very long contexts, SWA with attention sinks is recommended as the first try for fixed, low-memory scenarios without retraining.

Original post →

More from Research

Research channel →