Logical Reasoning on BBEH
58.9AccuracyVERGE Full
Evaluation results
| Method | Links | |
|---|---|---|
| VERGE Fullbackbone=GPT-120B2026.01 | 58.9 | |
| VERGE w/o MCSbackbone=GPT-120B2026.01 | 54.1 | |
| VERGE w/o Rtbackbone=GPT-120B2026.01 | 50.2 | |
| VERGE Fullbackbone=GPT-20B2026.01 | 49.9 | |
| VERGE Fullbackbone=Sonnet-3.72026.01 | 45.9 | |
| VERGE w/o Rtbackbone=GPT-20B2026.01 | 43.7 | |
| VERGE w/o Rtbackbone=Sonnet-3.72026.01 | 43.5 | |
| VERGE w/o MCSbackbone=GPT-20B2026.01 | 42.2 | |
| SCbackbone=GPT-120B2026.01 | 41.8 | |
| VERGE w/o MCSbackbone=Sonnet-3.72026.01 | 40.4 | |
| SCbackbone=GPT-20B2026.01 | 38.4 | |
| CoTbackbone=GPT-120B2026.01 | 38.4 | |
| SRbackbone=Sonnet-3.72026.01 | 37.8 | |
| SRbackbone=GPT-120B2026.01 | 37.4 | |
| Qwen2.5-32B-Instruct + Bootcamp-SFTModel=Qwen2.5-32B-Instruct, Training Stage=Bootcamp-SFT2025.08 | 36.7 | |
| Qwen2.5-32B-Instruct + Bootcamp-SFT-RLModel=Qwen2.5-32B-Instruct, Training Stage=Bootcamp-SFT-RL2025.08 | 35.9 | |
| CoTbackbone=GPT-20B2026.01 | 34.4 | |
| CoTbackbone=Sonnet-3.72026.01 | 33 | |
| SCbackbone=Sonnet-3.72026.01 | 29.4 | |
| SRbackbone=GPT-20B2026.01 | 28 | |
| DS-R1-Distilled-Qwen-32BModel=DS-R1-Distilled-Qwen-32B2025.08 | 27.8 | |
| DS-R1-Distilled-Qwen-32B + Bootcamp-RLModel=DS-R1-Distilled-Qwen-32B, Training Stage=Bootcamp-RL2025.08 | 27 | |
| DSBbackbone=Sonnet-3.72026.01 | 24 | |
| Qwen2.5-32B-Instruct + Bootcamp-RLModel=Qwen2.5-32B-Instruct, Training Stage=Bootcamp-RL2025.08 | 20.3 | |
| DSBbackbone=GPT-120B2026.01 | 20.2 | |
| Qwen2.5-32B-InstructModel=Qwen2.5-32B-Instruct2025.08 | 19.4 | |
| DSBbackbone=GPT-20B2026.01 | 10 |