Mathematical Reasoning on GSM8K (Accuracy, Pass@8)
97.8AccuracyQwen3-8B (Thinking)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-8B (Thinking)Model=Qwen3-8B, Variant=Thinking2026.05 | 97.8 | 98 | |
| Qwen3-1.7B (Thinking)Model=Qwen3-1.7B, Variant=Thinking2026.05 | 90.9 | 97 | |
| Qwen3-1.7B + Guided (ρ¯ ≈ 0.16)Model=Qwen3-1.7B, Intervention Rate (rho)=≈ 0.162026.05 | 88.9 | 98 | |
| Qwen3-0.6B (Thinking)Model=Qwen3-0.6B, Variant=Thinking2026.05 | 81.6 | 94 | |
| Qwen3-0.6B + Guided (ρ¯ ≈ 0.13)Model=Qwen3-0.6B, Intervention Rate (rho)=≈ 0.132026.05 | 80.4 | 99 | |
| Qwen3-1.7B-BaseModel=Qwen3-1.7B, Variant=Base2026.05 | 75.4 | 93 | |
| Qwen3-1.7B + Guided (ρ¯ ≈ 0.06)Model=Qwen3-1.7B, Intervention Rate (rho)=≈ 0.062026.05 | 72 | 97 | |
| Qwen3-0.6B + Guided (ρ¯ ≈ 0.04)Model=Qwen3-0.6B, Intervention Rate (rho)=≈ 0.042026.05 | 56.5 | 96 | |
| Qwen3-0.6B-BaseModel=Qwen3-0.6B, Variant=Base2026.05 | 29.9 | 82 |