List your company
ResearchFormerly WizWand

Mathematical Reasoning on Olympiad Bench (test)

83.53Pass@1 AccuracyEDS
14.338832.301950.26568.2281Feb 3, 2026Feb 13, 2026Feb 24, 2026Mar 6, 2026Mar 17, 2026Mar 27, 2026Apr 7, 2026
Updated 6mo ago

Evaluation results

MethodLinks
EDSBackbone=QWQ-32B2026.0283.538,710
BaseBackbone=QWQ-32B2026.0281.99,602
Full-suppressBackbone=QWQ-32B2026.0281.458,454
BaseBackbone=Qwen3-8B2026.0280.4210,480
EDSBackbone=Qwen3-8B2026.0279.829,739
Full-suppressBackbone=Qwen3-8B2026.0279.539,540
BaseBackbone=DeepSeek-7B2026.02697,913
EDSBackbone=DeepSeek-7B2026.0267.957,183
Full-suppressBackbone=DeepSeek-7B2026.0266.917,002
Evo-L2S-02Model Scale=7B, Merging Strategy=Multi-Objective Evolutionary Merging2026.0452.92,932.2
Evo-L2S-01Model Scale=7B, Merging Strategy=Multi-Objective Evolutionary Merging2026.0451.14,206.5
Evo-L2S (Accuracy)Model Scale=7B, Merging Strategy=Single-Objective Evolutionary Merging2026.0450.74,147
ACM-TAModel Scale=7B, Merging Strategy=Activation-informed Merging (ACM)2026.0450.22,292.2
Task ArithmeticModel Scale=7B, Merging Strategy=Arithmetic Merging2026.0449.34,304.3
TIES-MergingModel Scale=7B, Merging Strategy=Arithmetic Merging2026.0448.72,610.8
Average MergingModel Scale=7B, Merging Strategy=Arithmetic Merging2026.0447.92,223.1
DeepSeek-R1-Distill-Qwen-7BModel Scale=7B2026.0447.76,187.3
Evo-L2S-03Model Scale=7B, Merging Strategy=Multi-Objective Evolutionary Merging2026.0447.72,433.8
ACM-TIESModel Scale=7B, Merging Strategy=Activation-informed Merging (ACM)2026.04452,140.1
ACM-AverageModel Scale=7B, Merging Strategy=Activation-informed Merging (ACM)2026.0441.81,981.5
Evo-L2S-04Model Scale=7B, Merging Strategy=Multi-Objective Evolutionary Merging2026.0440.71,448.5
Evo-L2S (Length)Model Scale=7B, Merging Strategy=Single-Objective Evolutionary Merging2026.04371,295.7
Evo-L2S-05Model Scale=7B, Merging Strategy=Multi-Objective Evolutionary Merging2026.0436.91,283.2
Qwen2.5-Math-7BModel Scale=7B2026.04171,228.2