List your company
ResearchFormerly WizWand

Mathematical Reasoning on GSM8K v1.1 (test)

81.7AccuracyExpertCondenser
44.36454.05763.7573.443Apr 24, 2026
Updated 6mo ago

Evaluation results

MethodLinks
ExpertCondenserModel=GPT-OSS, Model Size=20B, Post-train Type=ExpertCondenser, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0481.7
DenseMixerModel=GPT-OSS, Model Size=20B, Post-train Type=DenseMixer, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0480.1
SFTModel=GPT-OSS, Model Size=20B, Post-train Type=SFT, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0478.2
ESFTModel=GPT-OSS, Model Size=20B, Post-train Type=ESFT, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0476.6
ExpertCondenserModel=OLMoE, Model Size=7B, Post-train Type=ExpertCondenser, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0468.4
ExpertCondenserModel=OLMoE, Model Size=7B, Post-train Type=ExpertCondenser, Post-training Dataset=math14k, Evaluation Protocol=zero-shot2026.0467.8
DenseMixerModel=OLMoE, Model Size=7B, Post-train Type=DenseMixer, Post-training Dataset=math14k, Evaluation Protocol=zero-shot2026.0466.2
SFTModel=OLMoE, Model Size=7B, Post-train Type=SFT, Post-training Dataset=math14k, Evaluation Protocol=zero-shot2026.0465.3
DenseMixerModel=OLMoE, Model Size=7B, Post-train Type=DenseMixer, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0464.6
ESFTModel=OLMoE, Model Size=7B, Post-train Type=ESFT, Post-training Dataset=math14k, Evaluation Protocol=zero-shot2026.0464.4
SFTModel=OLMoE, Model Size=7B, Post-train Type=SFT, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0463.6
ExpertCondenserModel=DeepSeek-V2-Lite, Model Size=16B, Post-train Type=ExpertCondenser, Post-training Dataset=math14k, Evaluation Protocol=zero-shot2026.0463.6
ESFTModel=OLMoE, Model Size=7B, Post-train Type=ESFT, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0462.2
DenseMixerModel=DeepSeek-V2-Lite, Model Size=16B, Post-train Type=DenseMixer, Post-training Dataset=math14k, Evaluation Protocol=zero-shot2026.0461.6
ExpertCondenserModel=DeepSeek-V2-Lite, Model Size=16B, Post-train Type=ExpertCondenser, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0459.4
ExpertCondenserModel=Qwen1.5-MoE, Model Size=14B, Post-train Type=ExpertCondenser, Post-training Dataset=math14k, Evaluation Protocol=zero-shot2026.0458.8
ESFTModel=DeepSeek-V2-Lite, Model Size=16B, Post-train Type=ESFT, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0458.6
SFTModel=DeepSeek-V2-Lite, Model Size=16B, Post-train Type=SFT, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0458.4
ESFTModel=DeepSeek-V2-Lite, Model Size=16B, Post-train Type=ESFT, Post-training Dataset=math14k, Evaluation Protocol=zero-shot2026.0458.2
DenseMixerModel=DeepSeek-V2-Lite, Model Size=16B, Post-train Type=DenseMixer, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0457.6
SFTModel=DeepSeek-V2-Lite, Model Size=16B, Post-train Type=SFT, Post-training Dataset=math14k, Evaluation Protocol=zero-shot2026.0457.6
ExpertCondenserModel=Qwen1.5-MoE, Model Size=14B, Post-train Type=ExpertCondenser, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0457.2
DenseMixerModel=Qwen1.5-MoE, Model Size=14B, Post-train Type=DenseMixer, Post-training Dataset=math14k, Evaluation Protocol=zero-shot2026.0455.8
DenseMixerModel=Qwen1.5-MoE, Model Size=14B, Post-train Type=DenseMixer, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0454.6
ESFTModel=Qwen1.5-MoE, Model Size=14B, Post-train Type=ESFT, Post-training Dataset=math14k, Evaluation Protocol=zero-shot2026.0452.5
SFTModel=Qwen1.5-MoE, Model Size=14B, Post-train Type=SFT, Post-training Dataset=math14k, Evaluation Protocol=zero-shot2026.0451.8
ESFTModel=Qwen1.5-MoE, Model Size=14B, Post-train Type=ESFT, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0446.9
SFTModel=Qwen1.5-MoE, Model Size=14B, Post-train Type=SFT, Post-training Dataset=math7k, Evaluation Protocol=zero-shot2026.0445.8