List your company
ResearchFormerly WizWand

Mathematical Reasoning on MATH (MATH, Avg.)

61.2MATH ScoreGrid
1.753617.186832.6248.0532May 20, 2026May 27, 2026Jun 3, 2026Jun 10, 2026Jun 17, 2026Jun 24, 2026Jul 1, 2026
Updated 4mo ago

Evaluation results

MethodLinks
GridBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.0761.274.27
CAUSALMIXBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.0760.5874.72
EqualBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.0756.7873.62
DMOBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.0754.3872
DoReMiBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.0742.2265.39
ODMBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.0741.1664.74
RegMixBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.0740.865.21
Qwen-4B (teacher)Setting=Teachers (reference), Few-shot protocol=3-shot2026.0527.7666.42
Phi-mini (teacher)Setting=Teachers (reference), Few-shot protocol=3-shot2026.0519.363.72
X-TokenSetting=Multi-teacher, Teachers=Phi-mini + Llama-3B, Few-shot protocol=3-shot2026.059.0240.48
Llama-3B (teacher)Setting=Teachers (reference), Few-shot protocol=3-shot2026.058.8246.93
X-TokenSetting=Multi-teacher, Teachers=Phi-mini + Qwen-4B + Llama-3B, Few-shot protocol=3-shot2026.058.5640.15
X-Token (H-KL)Setting=Cross tokenizer (single teacher), Teacher=Phi-mini, Few-shot protocol=3-shot2026.058.3239.18
Llama-3B → 1BSetting=Same tokenizer, Teacher=Llama-3B, Few-shot protocol=3-shot2026.058.1638.4
X-TokenSetting=Multi-teacher, Teachers=Phi-mini + Qwen-4B, Few-shot protocol=3-shot2026.058.138.49
X-Token (P-KL)Setting=Cross tokenizer (single teacher), Teacher=Qwen-4B, Few-shot protocol=3-shot2026.057.9638.85
GOLDSetting=Cross tokenizer (single teacher), Teacher=Phi-mini, Few-shot protocol=3-shot2026.057.838.66
Continued pre-trainingSetting=No distillation, Few-shot protocol=3-shot2026.056.936.63
ULDSetting=Cross tokenizer (single teacher), Teacher=Phi-mini, Few-shot protocol=3-shot2026.056.2438.31
Llama-1B (base)Setting=No distillation, Few-shot protocol=3-shot2026.055.4833.96
GOLDSetting=Cross tokenizer (single teacher), Teacher=Qwen-4B, Few-shot protocol=3-shot2026.054.535.03
ULDSetting=Cross tokenizer (single teacher), Teacher=Qwen-4B, Few-shot protocol=3-shot2026.054.0436.77