Hard LLM Reasoning on HLE
15.5AccuracyHyPER
Evaluation results
| Method | Links | ||
|---|---|---|---|
| HyPERModel=Qwen3-Next2026.02 | 15.5 | 0.76 | |
| HyPERModel=Qwen3-30B2026.02 | 13.5 | 0.81 | |
| DeepConfModel=Qwen3-30B2026.02 | 11.5 | 0.91 | |
| DeepConfModel=Qwen3-Next2026.02 | 11 | 0.84 | |
| RoEModel=Qwen3-30B2026.02 | 7 | 3.55 | |
| SCModel=Qwen3-Next2026.02 | 7 | 1 | |
| SCModel=Qwen3-30B2026.02 | 6.5 | 1 | |
| RoEModel=Qwen3-Next2026.02 | 6.5 | 2.98 | |
| Self-CertaintyModel=Qwen3-30B2026.02 | 2.5 | 1.15 | |
| Self-CertaintyModel=Qwen3-Next2026.02 | 2.5 | 1.25 |