Mathematical Reasoning on GSM8K 200 held-out questions
90.4AccuracyMAGE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MAGEBackbone=Qwen3-8B, Judge=Claude Sonnet 4.6, MAGE Executor Model=Opus2026.05 | 90.4 | 7.9 | |
| MAGEBackbone=Qwen3-8B, Judge=Claude Sonnet 4.6, MAGE Executor Model=Sonnet2026.05 | 88.2 | 7.9 | |
| 8-shotBackbone=Qwen3-8B, Judge=Claude Sonnet 4.6, Prompting Scaffold=8-shot2026.05 | 82.5 | — | |
| ReActBackbone=Qwen3-8B, Judge=Claude Sonnet 4.6, Prompting Scaffold=ReAct2026.05 | 80.1 | — | |
| SC10Backbone=Qwen3-8B, Judge=Claude Sonnet 4.6, Prompting Scaffold=SC10, Samples=102026.05 | 72 | — | |
| 0-shot CoTBackbone=Qwen3-8B, Judge=Claude Sonnet 4.6, Prompting Scaffold=0-shot CoT2026.05 | 64.5 | — | |
| ReflexionBackbone=Qwen3-8B, Judge=Claude Sonnet 4.6, Prompting Scaffold=Reflexion2026.05 | 63 | — |