Question Answering on ARC Challenge (Efficiency Metrics)
50.9ScoreSelf-Pruned KV
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Self-Pruned KVAttention Mechanism=Self-Pruned KV, tau=0.5, Model Size=8.1B, Training Context=32k2026.05 | 50.9 | 0.7 | 50.6 | |
| VanillaAttention Mechanism=Full Attention, Model Size=8.1B, Training Context=32k2026.05 | 50.6 | — | — |