method · first seen 22 Oct 2026
window attention
Also: windowed self-attention
Restricting self-attention computation to local groups of tokens to reduce cost.
Mentions per week
PapersCodeDiscussion
Earliest sources
- MixPT: Point Transformer with Interleaved Window and Global AttentionPapers · 22 Oct 2026 · Minkyun Seo, Hyungtae Lim, Jaesik Park
Latest sources
- MixPT: Point Transformer with Interleaved Window and Global AttentionPapers · 22 Oct 2026 · Minkyun Seo, Hyungtae Lim, Jaesik Park