Question Answering on GPQA Diamond (solve rate)
44.44Solve RateBifrost
Evaluation Results
| Method | Links | |
|---|---|---|
| BifrostBackbone=Llama-3.2 1B2026.02 | 44.44 | |
| BifrostBackbone=Llama-3.1 8B2026.02 | 37.88 | |
| ReflexionBackbone=Llama-3.1 8B2026.02 | 33.84 | |
| BifrostBackbone=Llama-3.2 3B2026.02 | 32.83 | |
| DoTBackbone=Llama-3.1 8B2026.02 | 30.81 | |
| ICLBackbone=Llama-3.2 3B2026.02 | 25.25 | |
| ICLBackbone=Llama-3.2 1B2026.02 | 23.23 | |
| ICLBackbone=Llama-3.1 8B2026.02 | 22.22 | |
| RISEBackbone=Llama-3.2 1B, Attempts=52026.02 | 21.21 | |
| ReflexionBackbone=Llama-3.2 3B2026.02 | 20.2 | |
| PaprikaBackbone=Llama-3.2 3B2026.02 | 20.2 | |
| RISEBackbone=Llama-3.2 3B, Attempts=52026.02 | 19.19 | |
| PaprikaBackbone=Llama-3.2 1B2026.02 | 18.69 | |
| CoTBackbone=Llama-3.1 8B2026.02 | 17.68 | |
| BoTBackbone=Llama-3.2 1B2026.02 | 16.67 | |
| DoTBackbone=Llama-3.2 3B2026.02 | 13.64 | |
| CoTBackbone=Llama-3.2 3B2026.02 | 12.63 | |
| RISEBackbone=Llama-3.1 8B, Attempts=52026.02 | 12.63 | |
| BoTBackbone=Llama-3.2 3B2026.02 | 12.12 | |
| PaprikaBackbone=Llama-3.1 8B2026.02 | 12.12 | |
| BoTBackbone=Llama-3.1 8B2026.02 | 11.11 | |
| RISEBackbone=Llama-3.2 1B, Attempts=12026.02 | 6.06 | |
| CoTBackbone=Llama-3.2 1B2026.02 | 4.04 | |
| RISEBackbone=Llama-3.2 3B, Attempts=12026.02 | 4.04 | |
| RISEBackbone=Llama-3.1 8B, Attempts=12026.02 | 3.54 | |
| ReflexionBackbone=Llama-3.2 1B2026.02 | 1.1 | |
| DoTBackbone=Llama-3.2 1B2026.02 | 1.01 |