List your company
ResearchFormerly WizWand

Mathematical Reasoning on GSM-Hard Natural Language version (test)

18.27AccuracyExplicit CoT
10.709212.672114.63516.5979Jun 30, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Explicit CoTBackbone=LLAMA-3.2-3B-INSTRUCT, Training Data=GSM8k-Aug (natural language)2026.0618.27
LOTUSBackbone=LLAMA-3.2-3B-INSTRUCT, Training Data=GSM8k-Aug (natural language)2026.0616.27
KaVaBackbone=LLAMA-3.2-3B-INSTRUCT, Training Data=GSM8k-Aug (natural language)2026.0614.8
CODIBackbone=LLAMA-3.2-3B-INSTRUCT, Training Data=GSM8k-Aug (natural language)2026.0613.6
PCCoTBackbone=LLAMA-3.2-3B-INSTRUCT, Training Data=GSM8k-Aug (natural language)2026.0611