← back to paper

Review history

arxiv: 2608.06776 · 2 revisions

Faster Query-Key Learning Sharpens Attention in Self-Attention Models

1. 2026-08-15 CONDITIONAL MODERATE v1.4.0-alltime-deepseek-medium novelty 6.0
   159224 ms 25232 in 18847 out 2026-08-15T14:30:17.476889+00:00
2. 2026-08-10 CONDITIONAL MODERATE v1.4.0-daily-deepseek-medium novelty 6.0
   133296 ms 25431 in 15856 out 2026-08-10T20:58:54.001848+00:00