List your company
ResearchFormerly WizWand

Mathematical Reasoning on AIME 2024 (Accuracy and Reasoning Analysis)

0.8148AccuracyQwen3-8B
0.2794080.4184040.55740.696396Mar 4, 2025Apr 30, 2025Jun 27, 2025Aug 23, 2025Oct 20, 2025Dec 16, 2025Feb 12, 2026
Updated 8mo ago

Evaluation results

MethodLinks
Qwen3-8BPruning strategy=80% lowest-entropy steps2025.080.814811,533.57
Qwen3-8BPruning strategy=None2025.080.793120,936.57
DeepSeek-R1-14BPruning strategy=None2025.080.655215,414.83
DeepSeek-R1-7BPruning strategy=None2025.080.633315,843.43
DeepSeek-R1-14BPruning strategy=80% lowest-entropy steps2025.080.58628,705.57
DeepSeek-R1-7BPruning strategy=80% lowest-entropy steps2025.080.566710,092.8
VanillaBase Model=DeepSeek-R1-Distill-Llama-8B2026.020.485410,774
ARLCPBase Model=DeepSeek-R1-Distill-Llama-8B2026.020.44587,436
ARLCPBase Model=Qwen3-1.7B2026.020.42929,466
VanillaBase Model=Qwen3-1.7B2026.020.387513,138
Qwen2.5-Math-72B-InstructEvaluating Model=Qwen2.5-Math-72B-Instruct2025.030.3—
DeepSeek-R1-Distill-Qwen-7BEvaluating Model=DeepSeek-R1-Distill-Qwen-7B2025.03—4,159