List your company

Competition Math Evaluation Suite (AIME, Brumo, HMMT)

71.2AIME ScoreParaGator-4B-Instruct-Principia + aggregation
5.6822.6939.756.71Mar 19, 2026

Evaluation results

MethodLinks
ParaGator-4B-Instruct-Principia + aggregationBackbone=Qwen3-4B-Instruct-2507, Aggregation=Enabled2026.0371.246.6566.2861.38
Online Multitask + aggregationBackbone=Qwen3-4B-Instruct-2507, Aggregation=Enabled2026.0364.2445.2667.3358.94
Dr.GRPO (with KL) + aggregationBackbone=Qwen3-4B-Instruct-2507, Aggregation=Enabled2026.0363.6547.2965.5258.79
Offline aggregation (AggLM) + aggregationBackbone=Qwen3-4B-Instruct-2507, Aggregation=Enabled2026.0362.4243.1665.1456.91
ParaGator-4B-Instruct-PrincipiaBackbone=Qwen3-4B-Instruct-2507, Aggregation=None2026.0359.4442.3259.6753.81
Online MultitaskBackbone=Qwen3-4B-Instruct-2507, Aggregation=None2026.0358.3641.7359.4553.18
Dr.GRPO (with KL)Backbone=Qwen3-4B-Instruct-2507, Aggregation=None2026.0358.0244.6961.5654.2
Qwen3-4B-Instruct-2507 + aggregationBackbone=Qwen3-4B-Instruct-2507, Aggregation=Enabled2026.0356.4635.9461.8851.43
Offline aggregation (AggLM)Backbone=Qwen3-4B-Instruct-2507, Aggregation=None2026.0351.0637.4754.9147.81
Qwen3-4B-Instruct-2507Backbone=Qwen3-4B-Instruct-2507, Aggregation=None2026.0347.7130.8355.4244.65
ParaGator-Zero-4B-Principia + aggregationBackbone=Qwen3-4B-Base, Aggregation=Enabled2026.0327.9631.06—23.72
Online Multitask + aggregationBackbone=Qwen3-4B-Base, Aggregation=Enabled2026.0326.1929.5414.1123.28
Dr.GRPO + aggregationBackbone=Qwen3-4B-Base, Aggregation=Enabled2026.0325.9531.0713.123.37
ParaGator-Zero-4B-PrincipiaBackbone=Qwen3-4B-Base, Aggregation=None2026.0321.0522.587.0416.89
Dr.GRPOBackbone=Qwen3-4B-Base, Aggregation=None2026.0320.9323.198.1417.42
Online MultitaskBackbone=Qwen3-4B-Base, Aggregation=None2026.0319.1422.377.9816.5
Offline aggregation (AggLM) + aggregationBackbone=Qwen3-4B-Base, Aggregation=Enabled2026.0316.9125.6512.418.32
Qwen3-4B-Base + aggregationBackbone=Qwen3-4B-Base, Aggregation=Enabled2026.0310.314.252.59.02
Offline aggregation (AggLM)Backbone=Qwen3-4B-Base, Aggregation=None2026.039.1317.117.4211.22
Qwen3-4B-BaseBackbone=Qwen3-4B-Base, Aggregation=None2026.038.216.673.29.36