List your company

Mathematical Reasoning on AIME 2025 (Pass@k)

73.1Pass@kGCPO
19.74833.59947.4561.301May 12, 2026

Evaluation results

MethodLinks
GCPOk=64, Backbone=Qwen3-4B2026.0573.1
GCPOk=32, Backbone=Qwen3-4B2026.0569.8
DQOk=64, Backbone=Qwen3-4B2026.0567.3
GCPOk=16, Backbone=Qwen3-4B2026.0566.4
DQOk=32, Backbone=Qwen3-4B2026.0564.1
GRPOk=64, Backbone=Qwen3-4B2026.0563.2
GCPOk=8, Backbone=Qwen3-4B2026.0562.7
DQOk=16, Backbone=Qwen3-4B2026.0561.2
GRPOk=32, Backbone=Qwen3-4B2026.0559.9
GCPOk=4, Backbone=Qwen3-4B2026.0558.9
GRPOk=16, Backbone=Qwen3-4B2026.0556.9
DQOk=8, Backbone=Qwen3-4B2026.0555.6
GCPOk=2, Backbone=Qwen3-4B2026.0553.1
Basek=64, Backbone=Qwen3-4B2026.0552.7
GRPOk=8, Backbone=Qwen3-4B2026.0551.7
DQOk=4, Backbone=Qwen3-4B2026.0551.5
Basek=32, Backbone=Qwen3-4B2026.0547.8
GRPOk=4, Backbone=Qwen3-4B2026.0546.8
DQOk=2, Backbone=Qwen3-4B2026.0544.8
Basek=16, Backbone=Qwen3-4B2026.0541.4
GRPOk=2, Backbone=Qwen3-4B2026.0541.2
Basek=8, Backbone=Qwen3-4B2026.0534.3
Basek=4, Backbone=Qwen3-4B2026.0527.7
Basek=2, Backbone=Qwen3-4B2026.0521.8