List your company

Mathematical Reasoning on MinervaMath (Conditional Local Accuracy, Overall Accuracy)

52.68AccuracyQwen2.5-Math-1.5B-Instruct
9.967221.056132.14543.2339Jun 3, 2026

Evaluation results

MethodLinks
Qwen2.5-Math-1.5B-InstructModel scale=1.5B, Evaluation mode=0-shot2026.0652.68——
DLLGModel scale=1.5B, Evaluation mode=0-shot2026.0651.79——
RouterDCModel scale=1.5B, Evaluation mode=0-shot2026.0651.34——
EmbedLLMModel scale=1.5B, Evaluation mode=0-shot2026.0651.34——
Entropy WeightingModel scale=1.5B, Evaluation mode=0-shot2026.0648.66——
UniTeModel scale=1.5B, Evaluation mode=0-shot2026.0645.98——
GaCModel scale=1.5B, Evaluation mode=0-shot2026.0645.09——
Pack of LLMsModel scale=1.5B, Evaluation mode=0-shot2026.0644.64——
Token Maj-VotingModel scale=1.5B, Evaluation mode=0-shot2026.0638.39——
Qwen2.5-1.5B-InstructModel scale=1.5B, Evaluation mode=0-shot2026.0630.36——
LinearModel scale=1.5B, Evaluation mode=0-shot2026.0625.45——
Qwen2.5-Coder-1.5B-InstructModel scale=1.5B, Evaluation mode=0-shot2026.0624.55——
SLERPModel scale=1.5B, Evaluation mode=0-shot2026.0622.32——
Task ArithmeticModel scale=1.5B, Evaluation mode=0-shot2026.0611.61——
Cloud LLM2025.09——80.9
Collaboration-Aware TuningModel=Qwen2.5-1.5B2025.09—13.633.5
Collaboration-Aware TuningModel=Llama-3.2-3B2025.09—13.436.8
OursModel=Qwen2.5-1.5B2025.09—24.140.8
OursModel=Llama-3.2-3B2025.09—27.743.4
Task-Tuning OnlyModel=Qwen2.5-1.5B2025.09—20.620.6
Task-Tuning OnlyModel=Llama-3.2-3B2025.09—19.119.1
Task-Tuning&Naive OffloadingModel=Qwen2.5-1.5B2025.09—2138.2
Task-Tuning&Naive OffloadingModel=Llama-3.2-3B2025.09—19.437.1
Task-Tuning&RouterModel=Qwen2.5-1.5B2025.09—21.836.8
Task-Tuning&RouterModel=Llama-3.2-3B2025.09—20.239.1