Lil: Less is Less When Applying Post-Training Sparse-Attention Algorithms in Long-Decode Stage

Add code
Jan 06, 2026

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: