List your company
ResearchFormerly WizWand

Mathematical Reasoning on GSM (Accuracy & Average)

92.16GSM AccuracyCRDS-W
12.23632.985553.73574.4845Feb 14, 2026Mar 3, 2026Mar 20, 2026Apr 7, 2026Apr 24, 2026May 11, 2026May 29, 2026
Updated 5mo ago

Evaluation results

MethodLinks
CRDS-WModel=Ling-mini-2.0 (16B), Base Model=Ling-mini-2.0, Truncation Length=2048, Data Selection Ratio=5%, Number of parallel runs=52026.0292.1683.86
RandomModel=Ling-mini-2.0 (16B), Base Model=Ling-mini-2.0, Truncation Length=2048, Data Selection Ratio=5%, Number of parallel runs=52026.0291.4882.55
CRDS-RModel=Ling-mini-2.0 (16B), Base Model=Ling-mini-2.0, Truncation Length=2048, Data Selection Ratio=5%, Number of parallel runs=52026.0291.0283.31
Mid PPLModel=Ling-mini-2.0 (16B), Base Model=Ling-mini-2.0, Truncation Length=2048, Data Selection Ratio=5%, Number of parallel runs=52026.0291.0182.49
Ling-mini-2.0Model=Ling-mini-2.0 (16B), Base Model=Ling-mini-2.0, Truncation Length=2048, Data Selection Ratio=100%, Number of parallel runs=52026.0290.7583.25
RDS+Model=Ling-mini-2.0 (16B), Base Model=Ling-mini-2.0, Truncation Length=2048, Data Selection Ratio=5%, Number of parallel runs=52026.0290.5582.28
LengthModel=Ling-mini-2.0 (16B), Base Model=Ling-mini-2.0, Truncation Length=2048, Data Selection Ratio=5%, Number of parallel runs=52026.0285.0574.53
MoDSBase Model=Llama-3-8B, Training Data Scale=15%2026.0561.25—
MADS8BBase Model=Llama-3-8B, Training Data Scale=15%, Instructor Model=Llama-3.1-8B-Instruct2026.0560.72—
InsTagBase Model=Llama-3-8B, Training Data Scale=15%2026.0560.65—
NUGGETSBase Model=Llama-3-8B, Training Data Scale=15%2026.0560.5—
ClusterClipBase Model=Llama-3-8B, Training Data Scale=15%2026.0560.2—
MADS3BBase Model=Llama-3-8B, Training Data Scale=15%, Instructor Model=Llama-3.2-3B-Instruct2026.0559.89—
SelectITBase Model=Llama-3-8B, Training Data Scale=15%2026.0559.74—
DEITABase Model=Llama-3-8B, Training Data Scale=15%2026.0558.76—
IFDBase Model=Llama-3-8B, Training Data Scale=15%2026.0558.53—
FullBase Model=Llama-3-8B, Training Data Scale=100%2026.0557.92—
MADS3BBase Model=Llama-2-7B, Training Data Scale=15%, Instructor Model=Llama-3.2-3B-Instruct2026.0519.71—
MADS8BBase Model=Llama-2-7B, Training Data Scale=15%, Instructor Model=Llama-3.1-8B-Instruct2026.0519.33—
ClusterClipBase Model=Llama-2-7B, Training Data Scale=15%2026.0518.87—
SelectITBase Model=Llama-2-7B, Training Data Scale=15%2026.0518.87—
InsTagBase Model=Llama-2-7B, Training Data Scale=15%2026.0518.87—
FullBase Model=Llama-2-7B, Training Data Scale=100%2026.0518.65—
MoDSBase Model=Llama-2-7B, Training Data Scale=15%2026.0517.21—
NUGGETSBase Model=Llama-2-7B, Training Data Scale=15%2026.0517.13—
DEITABase Model=Llama-2-7B, Training Data Scale=15%2026.0516.75—
IFDBase Model=Llama-2-7B, Training Data Scale=15%2026.0515.31—