Reading Comprehension on BoolQ (Score, Relative Change (Δ %), Avg. Gate Density)
73ScoreSelf-Pruned KV
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Self-Pruned KVAttention Mechanism=Self-Pruned KV, tau=0.5, Model Size=8.1B, Training Context=32k2026.05 | 73 | 3.7 | 25.2 | |
| VanillaAttention Mechanism=Full Attention, Model Size=8.1B, Training Context=32k2026.05 | 70.4 | — | — |