List your company
ResearchFormerly WizWand

Mathematical Reasoning on GSM8K (Single Metric)

89.92GSM8K AccuracyGemma-3-4B
27.707243.858660.0176.1614Feb 1, 2025Mar 23, 2025May 12, 2025Jul 1, 2025Aug 20, 2025Oct 9, 2025Nov 28, 2025
Updated 5mo ago

Evaluation results

MethodLinks
Gemma-3-4BTotal Parameters=4B, Active Parameters=4B, Trained Tokens=4T2025.1189.92
LFM2-8B-A1BTotal Parameters=8.3B, Active Parameters=1.5B, Trained Tokens=13T2025.1184.38
DUET-IFBackbone=Llama-3-8b-Instruct2025.0284.2
Granite-4.0-HTotal Parameters=7B, Active Parameters=1B, Trained Tokens=15T2025.1182.64
LFM2-2.6BTotal Parameters=2.6B, Active Parameters=2.6B, Trained Tokens=11T2025.1182.41
SmolLM3-3BTotal Parameters=3.1B, Active Parameters=3.1B, Trained Tokens=11T2025.1181.12
DUET-LESSBackbone=Llama-3-8b-Instruct2025.0280.5
ODMBackbone=Llama-3-8b-Instruct2025.0277.3
IF onlyBackbone=Llama-3-8b-Instruct, M=200002025.0276.9
AioliBackbone=Llama-3-8b-Instruct2025.0276.5
Llama-3.2-3BTotal Parameters=3.2B, Active Parameters=3.2B, Trained Tokens=9T2025.1175.21
Multi-FidBackbone=Llama-3-8b-Instruct2025.0273.9
Qwen3-4BTotal Parameters=4B, Active Parameters=4B, Trained Tokens=36T2025.1168.46
Uniform + more training tokensBackbone=Llama-3-8b-Instruct2025.0264.4
RandomBackbone=Llama-3-8b-Instruct, M=500002025.0264.3
Gemma-3-1B# Total Params=1B, # Trained Tokens=2T2025.1159.59
LFM2-1.2B# Total Params=1.2B, # Trained Tokens=11T2025.1158.3
Qwen3-1.7B# Total Params=1.7B, # Trained Tokens=36T2025.1151.4
LFM2-700M# Total Params=0.70B, # Trained Tokens=11T2025.1146.4
Qwen3-0.6B# Total Params=0.6B, # Trained Tokens=36T2025.1136.47
Llama-3.2-1B# Total Params=1.2B, # Trained Tokens=9T2025.1135.71
LFM2-350M# Total Params=0.35B, # Trained Tokens=11T2025.1130.1