List your company

Mathematical Reasoning on GSM8K (Normalized Performance)

1.026Normalized performanceIso-C
0.732720.808860.8850.96114Jun 15, 2026

Evaluation results

MethodLinks
Iso-CCategory=State-of-the-Art (+Many-shot), Backbone=Llama-3.2-3B2026.061.026
FedMergeCategory=Iterative / FL, Backbone=Llama-3.2-3B2026.060.923
ColD FusionCategory=Iterative / FL, Backbone=Llama-3.2-3B2026.060.923
TIESCategory=Conventional Method (+Many-shot), Backbone=Llama-3.2-3B2026.060.923
Iso-CTSCategory=State-of-the-Art (+Many-shot), Backbone=Llama-3.2-3B2026.060.897
ConsensusTACategory=Conventional Method (+Many-shot), Backbone=Llama-3.2-3B2026.060.872
METISCategory=Ours, Backbone=Llama-3.2-3B2026.060.872
q-FedAvgCategory=Iterative / FL, Backbone=Llama-3.2-3B2026.060.846
DARECategory=Conventional Method (+Many-shot), Backbone=Llama-3.2-3B2026.060.846
TSV-MCategory=State-of-the-Art (+Many-shot), Backbone=Llama-3.2-3B2026.060.821
TULU 3Category=Data-Mixing, Backbone=Llama-3.2-3B2026.060.821
TA + Subspace BoostingCategory=State-of-the-Art (+Many-shot), Backbone=Llama-3.2-3B2026.060.744
Task ArithmeticCategory=Conventional Method (+Many-shot), Backbone=Llama-3.2-3B2026.060.744