Sibling-model judgment on Math AMC12 + AIME24 25 + HMMTFeb2025
0.97AUROCGnosis-SelfJudge
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Gnosis-SelfJudgeJudged Model=Qwen3 8B-Hybrid2025.12 | 0.97 | 0.97 | 0.08 | |
| Gnosis-Qwen1.7B as RMJudged Model=Qwen3 8B-Hybrid2025.12 | 0.97 | 0.99 | 0.07 | |
| Gnosis-SelfJudgeJudged Model=Qwen3 4B-Thinking2025.12 | 0.96 | 0.98 | 0.05 | |
| Gnosis-SelfJudgeJudged Model=Qwen3 1.7B-Hybrid2025.12 | 0.95 | 0.95 | 0.09 | |
| Gnosis-Qwen1.7B as RMJudged Model=Qwen3 4B-Thinking2025.12 | 0.93 | 0.97 | 0.18 | |
| SkyworkRM-Qwen3-8BJudged Model=Qwen3 1.7B-Hybrid2025.12 | 0.9 | 0.92 | 0.14 | |
| SkyworkRM-Qwen3-8BJudged Model=Qwen3 4B-Thinking2025.12 | 0.89 | 0.94 | 0.22 | |
| SkyworkRM-Qwen3-8BJudged Model=Qwen3 8B-Hybrid2025.12 | 0.86 | 0.95 | 0.23 |