List your company

Mathematical Reasoning on OLYMPIAD (mean@8)

43.62mean@8Rollout-GDRO
22.24827.796533.34538.8935Jan 27, 2026

Evaluation results

MethodLinks
Rollout-GDROBackbone=Qwen3-8B-Base2026.0143.62
Prompt-GDROBackbone=Qwen3-8B-Base2026.0142.43
Prompt-GDROBackbone=Qwen3-4B-Base2026.0140.98
Rollout-GDROBackbone=Qwen3-4B-Base2026.0139.28
GRPO (Baseline)Backbone=Qwen3-8B-Base2026.0136.92
GRPO (Baseline)Backbone=Qwen3-4B-Base2026.0130.48
Rollout-GDROBackbone=Qwen3-1.7B-Base2026.0126.71
Prompt-GDROBackbone=Qwen3-1.7B-Base2026.0125.07
GRPO (Baseline)Backbone=Qwen3-1.7B-Base2026.0123.07