Math Problem Solving on GSM8K (solve rate)
90.22Solve RateBifrost
Evaluation Results
| Method | Links | |
|---|---|---|
| BifrostModel=Llama-3.1-8B2026.02 | 90.22 | |
| ReflexionModel=Llama-3.1-8B2026.02 | 87.41 | |
| RISE (5 attempts)Model=Llama-3.1-8B2026.02 | 86.13 | |
| BifrostModel=Llama-3.2-3B2026.02 | 81.41 | |
| ICLModel=Llama-3.1-8B2026.02 | 80.78 | |
| RISE (5 attempts)Model=Llama-3.2-3B2026.02 | 78.62 | |
| DoTModel=Llama-3.2-3B2026.02 | 76.8 | |
| DoTModel=Llama-3.1-8B2026.02 | 76.72 | |
| CoTModel=Llama-3.1-8B2026.02 | 74.65 | |
| CoTModel=Llama-3.2-3B2026.02 | 70.18 | |
| ICLModel=Llama-3.2-3B2026.02 | 62.52 | |
| ReflexionModel=Llama-3.2-3B2026.02 | 62.4 | |
| PaprikaModel=Llama-3.1-8B2026.02 | 55.27 | |
| RISE (1 attempt)Model=Llama-3.1-8B2026.02 | 48.29 | |
| BifrostModel=Llama-3.2-1B2026.02 | 44.66 | |
| CoTModel=Llama-3.2-1B2026.02 | 42.15 | |
| RISE (5 attempts)Model=Llama-3.2-1B2026.02 | 40.71 | |
| PaprikaModel=Llama-3.2-3B2026.02 | 40.26 | |
| BoTModel=Llama-3.1-8B2026.02 | 39.73 | |
| RISE (1 attempt)Model=Llama-3.2-3B2026.02 | 36.92 | |
| BoTModel=Llama-3.2-3B2026.02 | 33.74 | |
| ICLModel=Llama-3.2-1B2026.02 | 23.81 | |
| PaprikaModel=Llama-3.2-1B2026.02 | 21.08 | |
| ReflexionModel=Llama-3.2-1B2026.02 | 15.85 | |
| RISE (1 attempt)Model=Llama-3.2-1B2026.02 | 14.48 | |
| BoTModel=Llama-3.2-1B2026.02 | 9.62 | |
| DoTModel=Llama-3.2-1B2026.02 | 3.71 |