Math on AIME-Extend
52.67AccuracyAdaRAS
Evaluation Results
| Method | Links | |
|---|---|---|
| AdaRASCategory=Steering2026.01 | 52.67 | |
| ProbingCategory=Steering2026.01 | 48 | |
| CoTPrompting=Vanilla CoT, Base Model=Qwen3-1.7B2026.01 | 47.33 | |
| OpenReasoning-Nemotron-1.5BCategory=Post-training2026.01 | 47.33 | |
| OpenThinker-3-1.5BCategory=Post-training2026.01 | 42.67 | |
| DeepSeek-R1-Distill-Qwen-1.5BCategory=Post-training2026.01 | 21.33 |