List your company

Mathematical Reasoning on LightEval MATH-Hard Level 5 (val)

66AccuracyQwen3-4B-Thinking
49.3653.685862.32Feb 3, 2026

Evaluation results

MethodLinks
Qwen3-4B-Thinkingmodel_variant=Thinking2026.02664,913
SELFCESTdelegation=learned clones2026.0257932
Base Qwen3-4Bmodel_variant=Base2026.02501,124