Science Question Answering on ARC Easy (Accuracy & Token Cost)
86.9AccuracyHyPER
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HyPERModel=DeepSeek-V22026.02 | 86.9 | 0.56 | |
| HyPERModel=OLMoE2026.02 | 86.5 | 0.36 | |
| DeepConfModel=DeepSeek-V22026.02 | 86.2 | 0.31 | |
| DeepConfModel=OLMoE2026.02 | 84.7 | 0.25 | |
| Self-CertaintyModel=DeepSeek-V22026.02 | 84.5 | 1.04 | |
| SCModel=DeepSeek-V22026.02 | 84.3 | 1 | |
| RoEModel=DeepSeek-V22026.02 | 84.3 | 1.86 | |
| RoEModel=OLMoE2026.02 | 83.8 | 2.04 | |
| Self-CertaintyModel=OLMoE2026.02 | 82.2 | 0.88 | |
| SCModel=OLMoE2026.02 | 80.1 | 1 |