Sibling-model judgment on TriviaQA
0.89AUROCGnosis-SelfJudge
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Gnosis-SelfJudgeJudged Model=Qwen3 4B-Thinking2025.12 | 0.89 | 0.89 | 0.05 | |
| Gnosis-SelfJudgeJudged Model=Qwen3 1.7B-Hybrid2025.12 | 0.87 | 0.79 | 0.1 | |
| Gnosis-SelfJudgeJudged Model=Qwen3 8B-Hybrid2025.12 | 0.86 | 0.9 | 0.09 | |
| Gnosis-Qwen1.7B as RMJudged Model=Qwen3 4B-Thinking2025.12 | 0.86 | 0.86 | 0.04 | |
| SkyworkRM-Qwen3-8BJudged Model=Qwen3 1.7B-Hybrid2025.12 | 0.84 | 0.73 | 0.23 | |
| Gnosis-Qwen1.7B as RMJudged Model=Qwen3 8B-Hybrid2025.12 | 0.84 | 0.88 | 0.12 | |
| SkyworkRM-Qwen3-8BJudged Model=Qwen3 4B-Thinking2025.12 | 0.73 | 0.89 | 0.17 | |
| SkyworkRM-Qwen3-8BJudged Model=Qwen3 8B-Hybrid2025.12 | 0.72 | 0.78 | 0.32 |