Mathematical Reasoning on AMC 23 (Avg@16)
69.1Avg@16HEAPA
Evaluation Results
| Method | Links | |
|---|---|---|
| HEAPABackbone=Qwen3-8B2026.01 | 69.1 | |
| GRPO (w. PS)Backbone=Qwen3-8B2026.01 | 68.6 | |
| GRPOBackbone=Qwen3-8B2026.01 | 68 | |
| GRPOBackbone=Qwen3-4B2026.01 | 58.9 | |
| GRPO (w. PS)Backbone=Qwen3-4B2026.01 | 58.7 | |
| HEAPABackbone=Qwen3-4B2026.01 | 58.6 | |
| GRPO (w. PS)Backbone=Qwen3-1.7B2026.01 | 40.6 | |
| HEAPABackbone=Qwen3-1.7B2026.01 | 40.3 | |
| GRPOBackbone=Qwen3-1.7B2026.01 | 39.8 | |
| HEAPABackbone=Qwen3-0.6B2026.01 | 31.9 | |
| GRPO (w. PS)Backbone=Qwen3-0.6B2026.01 | 30.6 | |
| GRPOBackbone=Qwen3-0.6B2026.01 | 28.7 |