List your company

Mathematical Reasoning on MINERVA (Pass@k)

38.4Pass@kGCPO
26.75229.77632.835.824May 12, 2026

Evaluation results

MethodLinks
GCPOk=64, Backbone=Qwen3-4B2026.0538.4
GCPOk=32, Backbone=Qwen3-4B2026.0537.8
DQOk=64, Backbone=Qwen3-4B2026.0537.7
GRPOk=64, Backbone=Qwen3-4B2026.0537.6
Basek=64, Backbone=Qwen3-4B2026.0537.4
DQOk=32, Backbone=Qwen3-4B2026.0537.3
GRPOk=32, Backbone=Qwen3-4B2026.0536.9
GCPOk=16, Backbone=Qwen3-4B2026.0536.2
DQOk=16, Backbone=Qwen3-4B2026.0535.7
Basek=32, Backbone=Qwen3-4B2026.0535.7
GRPOk=16, Backbone=Qwen3-4B2026.0535.3
Basek=16, Backbone=Qwen3-4B2026.0535.1
GCPOk=8, Backbone=Qwen3-4B2026.0534.9
DQOk=8, Backbone=Qwen3-4B2026.0534.4
GRPOk=8, Backbone=Qwen3-4B2026.0534.1
GCPOk=4, Backbone=Qwen3-4B2026.0533.2
DQOk=4, Backbone=Qwen3-4B2026.0532.9
Basek=8, Backbone=Qwen3-4B2026.0532.9
GRPOk=4, Backbone=Qwen3-4B2026.0532.6
GCPOk=2, Backbone=Qwen3-4B2026.0531
DQOk=2, Backbone=Qwen3-4B2026.0530.7
Basek=4, Backbone=Qwen3-4B2026.0530.6
GRPOk=2, Backbone=Qwen3-4B2026.0530.4
Basek=2, Backbone=Qwen3-4B2026.0527.2