Physical Commonsense Reasoning on PIQA (Score, Relative Change, and Gate Density)
0.806PIQA ScoreVanilla
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| VanillaAttention Mechanism=Full Attention, Model Size=8.1B, Training Context=32k2026.05 | 0.806 | — | — | |
| Self-Pruned KVAttention Mechanism=Self-Pruned KV, tau=0.5, Model Size=8.1B, Training Context=32k2026.05 | 0.803 | -30 | 37.2 |