List your company

Mathematical Reasoning on HMMT 2026

49AccuracySeirênes
0.84813.34925.8538.351May 12, 2026

Evaluation results

MethodLinks
SeirênesBackbone=Qwen3-30B-A3B, same budget=true2026.0549
DAPOBackbone=Qwen3-30B-A3B, same budget=true2026.0542.3
Qwen3-30B-A3BBackbone=Qwen3-30B-A3B2026.0541
SeirênesBackbone=Qwen3-4B-Instruct, same budget=true2026.0539.4
InTBackbone=Qwen3-4B-Instruct2026.0537.9
Dr.GRPOBackbone=Qwen3-4B-Instruct, same budget=true2026.0537.2
DAPOBackbone=Qwen3-4B-Instruct, same budget=true2026.0533
Scaf-GRPOBackbone=Qwen3-4B-Instruct2026.0532.2
Qwen3-4B-InstructBackbone=Qwen3-4B-Instruct2026.0531.9
SAGEBackbone=Qwen3-4B-Instruct2026.0529.9
SeirênesBackbone=Qwen2.5-7B-Instruct, same budget=true2026.058.5
LUFFYBackbone=Qwen2.5-7B-Instruct2026.056.3
SAGEBackbone=Qwen2.5-7B-Instruct2026.056.3
ExGRPOBackbone=Qwen2.5-7B-Instruct2026.056
DAPOBackbone=Qwen2.5-7B-Instruct, same budget=true2026.054.3
Qwen2.5-7B-InstructBackbone=Qwen2.5-7B-Instruct2026.052.7