List your company

Mathematical Reasoning on AMC 2024

70AccuracyGRPO
9.763225.401641.0456.6784May 9, 2026May 13, 2026May 17, 2026May 21, 2026May 25, 2026May 29, 2026Jun 2, 2026

Evaluation results

MethodLinks
GRPObase_model=Qwen2.5-Math-7B2026.0570
CoDistill-GRPObase_model=Qwen2.5-Math-7B, alpha=1, T=0, CT=true2026.0569.5
Qwen3 + DAPO + Hidden-AlignSize=14B2026.0665.62
Qwen3 + DAPOSize=14B2026.0664.1
CoDistill-GRPObase_model=Qwen2.5-Math-7B, alpha=1, T=0, CT=false2026.0564
CoDistill-GRPObase_model=Qwen2.5-Math-7B, alpha=1, T=50, CT=true2026.0563.5
CoDistill-GRPObase_model=Qwen2.5-Math-7B, alpha=2, T=0, CT=true2026.0563
Qwen3 + DAPO + Hidden-AlignSize=4B2026.0662.15
CoDistill-GRPObase_model=Qwen2.5-Math-7B, alpha=1, T=50, CT=false2026.0562
CoDistill-GRPObase_model=Qwen2.5-Math-7B, alpha=2, T=0, CT=false2026.0562
DAPO w/ EDASBackbone=Qwen3-8B2026.0559.38
Qwen3 + DAPOSize=4B2026.0658.75
DAPO w/ EDASBackbone=Qwen3-4B2026.0558.19
DAPOBackbone=Qwen3-4B2026.0557.15
DAPOBackbone=Qwen3-8B2026.0556.88
DAPO w/ En AdvBackbone=Qwen3-8B2026.0555.62
DAPO w/ En AdvBackbone=Qwen3-4B2026.0549.58
Qwen3 (base)Size=14B2026.0649.58
GRPO w/ EDASBackbone=Qwen3-8B2026.0549.38
GRPOBackbone=Qwen3-8B2026.0548.75
DAPO w/ EDASBackbone=Qwen3-4B-Base2026.0548.68
Qwen3 + DAPO + Hidden-AlignSize=1.7B2026.0647.57
Qwen3 + DAPOSize=1.7B2026.0647.43
Qwen2.5-Math-7Bstatus=Base Model2026.0547
Qwen3 (base)Size=4B2026.0645.35
PKPOBackbone=Qwen3-8B2026.0545.07
BaselineBackbone=Qwen3-4B2026.0544.31
BaselineBackbone=Qwen3-8B2026.0544.24
DAPO w/ En AdvBackbone=Qwen3-4B-Base2026.0534.24
DAPOBackbone=Qwen3-4B-Base2026.0533.68
Qwen3 (base)Size=1.7B2026.0627.99
BaselineBackbone=Qwen3-4B-Base2026.0512.08