List your company

Mathematical Reasoning on AIME 1983-2024 (val)

81AccuracyQwen3-4B-Thinking
45.6454.826473.18Feb 3, 2026

Evaluation results

MethodLinks
Qwen3-4B-Thinkingmodel_variant=Thinking2026.0281—
Base Qwen3-4Bmodel_variant=Base2026.02473,928
SELFCESTdelegation=learned clones2026.02471,850