Mathematical Problem Solving on AIME 2025 (Accuracy avg@k)
73.54Accuracy (avg@k)TRACE (FKL on Ky)
Evaluation Results
| Method | Links | |
|---|---|---|
| TRACE (FKL on Ky)Backbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 73.54 | |
| TRACE (RKL on Ey)Backbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 70.42 | |
| GRPOBackbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 68.96 | |
| RLSDBackbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 68.54 | |
| BaseBackbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 67.5 | |
| SRPOBackbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 54.17 | |
| SDPOBackbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 41.25 | |
| TRACE (RKL on Ey)Backbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 38.33 | |
| TRACE (FKL on Ky)Backbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 36.3 | |
| BaseBackbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 35 | |
| GRPOBackbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 35 | |
| RLSDBackbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 32.92 | |
| SRPOBackbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 23.75 | |
| SDPOBackbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 17.5 |