List your company
ResearchFormerly WizWand

Mathematical Reasoning on Minerva (pass@1, standard deviation)

45.96Pass@1MIPU
10.225619.502828.7838.0572May 25, 2025Jul 30, 2025Oct 5, 2025Dec 10, 2025Feb 15, 2026Apr 22, 2026Jun 28, 2026
Updated 4mo ago

Evaluation results

MethodLinks
MIPUModel=Qwen3-4B, Rollout Quantization=FP82026.0645.96—
MISModel=Qwen3-4B, Rollout Quantization=FP82026.0644.12—
LR-decayModel=Qwen3-4B, Rollout Quantization=FP82026.0643.75—
GRPO BaselineModel=Qwen3-4B, Rollout Quantization=FP82026.0643.39—
MISModel=Qwen3-1.7B, Rollout Quantization=FP82026.0634.19—
MIPUModel=Qwen3-1.7B, Rollout Quantization=FP82026.0633.82—
GRPO BaselineModel=Qwen3-1.7B, Rollout Quantization=FP82026.0631.68—
LR-decayModel=Qwen3-1.7B, Rollout Quantization=FP82026.0628.68—
UniRBackbone=Qwen2.5-3B, Trained Model=1.5B, Zero-shot=true, Temperature=0.62025.0526.3—
GRPO FullBackbone=Qwen2.5-3B, Trained Model=3B, Zero-shot=true, Temperature=0.62025.0525—
GRPO LoRABackbone=Qwen2.5-3B, Trained Model=3B, Zero-shot=true, Temperature=0.62025.0524.7—
Backbone + 1.5BBackbone=Qwen2.5-3B, Zero-shot=true, Temperature=0.62025.0520.7—
DeepSeekMath-7B-RLTrained Model=7B, Zero-shot=true, Temperature=0.62025.0520.6—
Backbone onlyBackbone=Qwen2.5-3B, Zero-shot=true, Temperature=0.62025.0519.2—
NuminaMath-7B-CoTTrained Model=7B, Zero-shot=true, Temperature=0.62025.0519.1—
Internlm2-math-plus-7BTrained Model=7B, Zero-shot=true, Temperature=0.62025.0517.3—
UniRBackbone=Llama3.2-3B, Trained Model=1B, Zero-shot=true, Temperature=0.62025.0516.6—
Mathstral-7B-v0.1Trained Model=7B, Zero-shot=true, Temperature=0.62025.0516.2—
Backbone + 1BBackbone=Llama3.2-3B, Zero-shot=true, Temperature=0.62025.0513.1—
Backbone onlyBackbone=Llama3.2-3B, Zero-shot=true, Temperature=0.62025.0512.5—
GRPO LoRABackbone=Llama3.2-3B, Trained Model=3B, Zero-shot=true, Temperature=0.62025.0512—
GRPO FullBackbone=Llama3.2-3B, Trained Model=3B, Zero-shot=true, Temperature=0.62025.0511.6—