inference/research
verifiedhigh · 0.90updated 2026-10-03

FlashAttention-4: Algorithm and Kernel Pipelining Co-Design for Asymmetric Hardware Scaling

paper

Fields

id
sources/pap-attn-flashattention-4
URL
https://arxiv.org/abs/2603.05451
Publisher
arXiv
Kind
paper
Published
2026-03-05
Accessed
2026-10-03
notes
arXiv preprint; no journal_ref. Authors overlap with the FlashAttention-3 team (Shah, Thakkar, Dao).

Notes

arXiv preprint; no journal_ref. Authors overlap with the FlashAttention-3 team (Shah, Thakkar, Dao).

Cross-references

This record references no other records.