List your company
ResearchFormerly WizWand

Arithmetic Reasoning on MATH

71AccuracyLLM-PeerReview-W
-2.8416.3335.554.67Jun 3, 2025Jul 7, 2025Aug 11, 2025Sep 15, 2025Oct 20, 2025Nov 24, 2025Dec 29, 2025
Updated 5mo ago

Evaluation results

MethodLinks
LLM-PeerReview-WType=LLM Ensemble2025.127167.8—
LLM-PeerReviewType=LLM Ensemble2025.1269.567.4—
Qwen2.5-7B-InstructType=Single LLM2025.1269.362.7—
Llama-3.1-8B-SelectionType=Our variants2025.1268.866.4—
Mistral-7B-SelectionType=Our variants2025.1266.464.7—
Qwen2.5-7B-SelectionType=Our variants2025.1266.265.5—
Smoothie-LocalType=LLM Ensemble2025.1261.859.8—
Qwen2-7B-SelectionType=Our variants2025.1261.762.1—
Agent-ForestType=LLM Ensemble2025.126160.1—
Qwen2-7B-InstructType=Single LLM2025.1259.856.6—
Smoothie-GlobalType=LLM Ensemble2025.1259.860.5—
GaCType=LLM Ensemble2025.125460.2—
Llama-3.1-8B-InstructType=Single LLM2025.1252.353.5—
RandomType=LLM Ensemble2025.1252.254.2—
Theoretical averageType=Single LLM2025.1251.954.1—
HiRABackbone Model=Gemma-2-9B, #Params=108.04M2025.0938.93——
ABBABackbone Model=Gemma-2-9B, #Params=108.04M2025.0938.5——
BoHABackbone Model=Gemma-2-9B, #Params=108.04M2025.0938.44——
FFTBackbone Model=Gemma-2-9B, #Params=9.24B2025.0938.02——
DoRABackbone Model=Gemma-2-9B, #Params=109.89M2025.0937.76——
PiSSABackbone Model=Gemma-2-9B, #Params=108.04M2025.0937.04——
rsLoRABackbone Model=Gemma-2-9B, #Params=108.04M2025.0936.88——
LoRABackbone Model=Gemma-2-9B, #Params=108.04M2025.0934.68——
Mistral-7B-InstructType=Single LLM2025.1226.543.5—
DiaBlor/N=N = 32, #Params=2.09%, Base Model=Llama2-7B, Evaluation Protocol=Fine-tuning2025.0620.4——
Full FTr/N=N/A, #Params=100%, Base Model=Llama2-7B, Evaluation Protocol=Fine-tuning2025.0619.8——
BoHABackbone Model=Mistral-7B, #Params=83.89M2025.0917.97——
ABBABackbone Model=Mistral-7B, #Params=83.89M2025.0917.96——
DoRABackbone Model=Mistral-7B, #Params=85.26M2025.0917.9——
MiLoRAr/N=r = 64, #Params=1.67%, Base Model=Llama2-7B, Evaluation Protocol=Fine-tuning2025.0617.8——
FFTBackbone Model=Mistral-7B, #Params=7.24B2025.0917.65——
DiaBlor/N=N = 64, #Params=1.04%, Base Model=Llama2-7B, Evaluation Protocol=Fine-tuning2025.0617.6——
HiRABackbone Model=Mistral-7B, #Params=83.89M2025.0917.31——
LoRAr/N=r = 64, #Params=1.67%, Base Model=Llama2-7B, Evaluation Protocol=Fine-tuning2025.0616.9——
PiSSABackbone Model=Mistral-7B, #Params=83.88M2025.0916.52——
rsLoRABackbone Model=Mistral-7B, #Params=83.88M2025.0916.24——
PiSSAr/N=r = 64, #Params=1.67%, Base Model=Llama2-7B, Evaluation Protocol=Fine-tuning2025.0615.8——
LoRABackbone Model=Mistral-7B, #Params=83.89M2025.0915.28——
Zero-shotr/N=N/A, #Params=0%, Base Model=Llama2-7B, Evaluation Protocol=Zero-shot2025.060——
ABBABackbone=Mistral-7B, #Params=83.89M2025.09——18.16
ABBABackbone=Gemma-2-9B, #Params=108.04M2025.09——38.82
BoHABackbone=Mistral-7B, #Params=83.89M2025.09——19.98
BoHABackbone=Gemma-2-9B, #Params=108.04M2025.09——38.58
FFTBackbone=Mistral-7B, #Params=7.24B2025.09——18.34
FFTBackbone=Gemma-2-9B, #Params=9.24B2025.09——39.44
HiRABackbone=Mistral-7B, #Params=83.89M2025.09——18.76
HiRABackbone=Gemma-2-9B, #Params=108.04M2025.09——40.02
LoRABackbone=Mistral-7B, #Params=83.89M2025.09——16.92
LoRABackbone=Gemma-2-9B, #Params=108.04M2025.09——34.84