List your company
ResearchFormerly WizWand

Mathematical Reasoning on GSM Hard (Accuracy)

68.6AccuracyRM-Regen
13.4827.7942.156.41Jan 28, 2026Feb 14, 2026Mar 4, 2026Mar 22, 2026Apr 9, 2026Apr 27, 2026May 15, 2026
Updated 5mo ago

Evaluation results

MethodLinks
RM-RegenModel=GPT-3.5, Verification Setting=Oracle2026.0368.6
RM-RegenModel=GPT-3.52026.0364
RM-RegenModel=Gemma2-9B, Verification Setting=Oracle2026.0353
Reflexion(3 iters)Model=Gemma2-9B, Verification Setting=Oracle2026.0350.8
RM-RegenModel=Gemma2-9B2026.0349.4
ProCoModel=Gemma2-9B, iterations=32026.0348.6
Best-of-N (N=3)Model=Gemma2-9B, Verification Setting=Oracle2026.0348.4
Best-of-N (N=3)Model=GPT-3.5, Verification Setting=Oracle2026.0346.8
ST CoTModel=Gemma2-9B, iterations=32026.0346.2
Reflexion(3 iters)Model=GPT-3.5, Verification Setting=Oracle2026.0345.2
ReflectEvoModel=Gemma2-9B, Verification Setting=Oracle2026.0344.4
RM-RegenModel=Llama3.1-8B, Verification Setting=Oracle2026.0341
ST CoTModel=GPT-3.5, iterations=32026.0339.8
ProCoModel=GPT-3.5, iterations=32026.0339.6
NEXATrained on=GSM8K, Evaluation Protocol=Without retraining2026.0537.13
Reflexion(3 iters)Model=Llama3.1-8B, Verification Setting=Oracle2026.0336.2
CoTTrained on=GSM8K, Evaluation Protocol=Without retraining2026.0536.07
RM-RegenModel=Llama3.1-8B2026.0334.8
SingleTrained on=GSM8K, Evaluation Protocol=Without retraining2026.0534.4
Best-of-N (N=3)Model=Llama3.1-8B, Verification Setting=Oracle2026.0334.2
Self-RefineModel=GPT-3.5, iterations=32026.0332.8
Self-RefineModel=Gemma2-9B, iterations=32026.0332.8
ReflectEvoModel=Llama3.1-8B, Verification Setting=Oracle2026.0332
ST CoTModel=Llama3.1-8B, iterations=32026.0331.8
ProCoModel=Llama3.1-8B, iterations=32026.0331.2
ALIGNReasoning Strategy=ALIGN2026.0124.6
SCReasoning Strategy=CoT-SC@maj162026.0124.1
Self-RefineModel=Llama3.1-8B, iterations=32026.0323
rStarReasoning Strategy=rStar2026.0122.8
FS-CoTReasoning Strategy=Few-shot CoT2026.0117.7
PrincipalReasoning Strategy=Principal2026.0115.6