List your company

Mathematical Problem Solving on MINERVA (Pass@k)

37.5Pass@kGCPO
19.40424.10228.833.498May 12, 2026

Evaluation results

MethodLinks
GCPOk=64, Base Model=Qwen3-1.7B2026.0537.5
DQOk=64, Base Model=Qwen3-1.7B2026.0537
GRPOk=64, Base Model=Qwen3-1.7B2026.0536.6
GCPOk=32, Base Model=Qwen3-1.7B2026.0536.4
DQOk=32, Base Model=Qwen3-1.7B2026.0535.8
GRPOk=32, Base Model=Qwen3-1.7B2026.0535.1
GCPOk=16, Base Model=Qwen3-1.7B2026.0534.6
DQOk=16, Base Model=Qwen3-1.7B2026.0534.1
GRPOk=16, Base Model=Qwen3-1.7B2026.0533.2
GCPOk=8, Base Model=Qwen3-1.7B2026.0532.8
Basek=64, Base Model=Qwen3-1.7B2026.0532
DQOk=8, Base Model=Qwen3-1.7B2026.0531.9
GRPOk=8, Base Model=Qwen3-1.7B2026.0531.5
GCPOk=4, Base Model=Qwen3-1.7B2026.0530
Basek=32, Base Model=Qwen3-1.7B2026.0529.7
DQOk=4, Base Model=Qwen3-1.7B2026.0529.3
GRPOk=4, Base Model=Qwen3-1.7B2026.0528.6
Basek=16, Base Model=Qwen3-1.7B2026.0527.6
GCPOk=2, Base Model=Qwen3-1.7B2026.0526.9
DQOk=2, Base Model=Qwen3-1.7B2026.0526.6
GRPOk=2, Base Model=Qwen3-1.7B2026.0525.7
Basek=8, Base Model=Qwen3-1.7B2026.0525.4
Basek=4, Base Model=Qwen3-1.7B2026.0522.5
Basek=2, Base Model=Qwen3-1.7B2026.0520.1