CARVE: Content-Aware Recurrent with Value Efficiency for Chunk-Parallel Linear Attention

Add code
Jun 25, 2026

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: