Mathematical Reasoning on GSM8k (Acc., log C)
88.7AccuracyBoBN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BoBNN=16, Backbone=tulu3-8b2026.07 | 88.7 | — | |
| BoBNN=4, Backbone=tulu3-8b2026.07 | 86.5 | — | |
| BoBNN=1, Backbone=tulu3-8b2026.07 | 85.4 | 0.43 | |
| pi_refN=16, Backbone=tulu3-8b2026.07 | 78.3 | — | |
| URIALN=16, Backbone=tulu3-8b2026.07 | 73.3 | — | |
| URIALN=1, Backbone=tulu3-8b2026.07 | 70.6 | 0.46 | |
| pi_refN=1, Backbone=tulu3-8b2026.07 | 69.6 | 0.47 | |
| BoBNN=16, Backbone=zephyr-7b2026.07 | 58.7 | — | |
| BoBNN=4, Backbone=zephyr-7b2026.07 | 44.5 | — | |
| pi_refN=16, Backbone=zephyr-7b2026.07 | 43.9 | — | |
| URIALN=16, Backbone=zephyr-7b2026.07 | 41.6 | — | |
| BoBNN=1, Backbone=zephyr-7b2026.07 | 31.5 | 0.63 | |
| pi_refN=1, Backbone=zephyr-7b2026.07 | 21.4 | 0.72 | |
| URIALN=1, Backbone=zephyr-7b2026.07 | 20.9 | 0.73 |