Code Reasoning on CRUX (Score Metric)
25.75CRUX ScoreHiLS-Attn NoPE-Q-Cal
Evaluation Results
| Method | Links | |
|---|---|---|
| HiLS-Attn NoPE-Q-CalTuning Protocol=Full param. CPT2026.07 | 25.75 | |
| HiLS-Attn RoPE-Q-CalTuning Protocol=Full param. CPT2026.07 | 25.62 | |
| HiLS-Attn HoPE-Q-CalTuning Protocol=Full param. CPT2026.07 | 25.12 | |
| Olmo3 BaseTuning Protocol=Freezing param.2026.07 | 24.62 | |
| LMK token tuningTuning Protocol=Freezing param.2026.07 | 24.62 | |
| Olmo3 512swaTuning Protocol=Full param. CPT2026.07 | 24.5 |