Mathematical Problem Solving on MATH 500 (Accuracy (avg@k))
97.83Accuracy (avg@k)TRACE (FKL on Ky)
Evaluation Results
| Method | Links | |
|---|---|---|
| TRACE (FKL on Ky)Backbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 97.83 | |
| TRACE (RKL on Ey)Backbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 97.48 | |
| GRPOBackbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 97.3 | |
| BaseBackbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 96.8 | |
| RLSDBackbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 96.68 | |
| SRPOBackbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 96.48 | |
| SDPOBackbone=Qwen3-8B, Symmetric Think evaluation=true2026.05 | 95.13 | |
| TRACE (RKL on Ey)Backbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 92.05 | |
| BaseBackbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 91.35 | |
| RLSDBackbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 91.12 | |
| TRACE (FKL on Ky)Backbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 91 | |
| GRPOBackbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 90.33 | |
| SRPOBackbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 86.02 | |
| SDPOBackbone=Qwen3-1.7B, Symmetric Think evaluation=true2026.05 | 81.73 |