List your company
ResearchFormerly WizWand

Mathematical Reasoning on MATH 500 (Accuracy, Tokens)

97.4AccuracyThoughtFold
83.46487.08290.794.318Jun 2, 2026
Updated 5mo ago

Evaluation results

MethodLinks
ThoughtFoldModel=Qwen3-8B2026.0697.42,933
ThoughtFoldModel=Qwen3-14B2026.0697.12,577
Short-RLModel=Qwen3-14B2026.0696.42,789
S-GRPOModel=Qwen3-14B2026.0696.42,652
GRPOModel=Qwen3-14B2026.0695.85,140
RL + Length PenaltyModel=Qwen3-14B2026.0695.82,866
S-GRPOModel=Qwen3-8B2026.0695.23,166
VanillaModel=Qwen3-14B2026.0695.25,078
ThoughtFoldModel=DeepSeek-R1-Distill-Qwen-7B2026.0694.42,089
GRPOModel=Qwen3-8B2026.0694.45,440
ThoughtFoldModel=DeepSeek-R1-Distill-Qwen-14B2026.0694.21,956
RL + Length PenaltyModel=Qwen3-8B2026.0694.23,247
GRPOModel=DeepSeek-R1-Distill-Qwen-7B2026.0693.65,317
S-GRPOModel=DeepSeek-R1-Distill-Qwen-14B2026.0693.62,146
VanillaModel=DeepSeek-R1-Distill-Qwen-14B2026.0693.53,844
VanillaModel=Qwen3-8B2026.0693.45,577
Short-RLModel=Qwen3-8B2026.0693.23,048
Short-RLModel=DeepSeek-R1-Distill-Qwen-14B2026.0693.12,021
S-GRPOModel=DeepSeek-R1-Distill-Qwen-7B2026.0692.42,252
RL + Length PenaltyModel=DeepSeek-R1-Distill-Qwen-14B2026.0692.41,993
RL + Length PenaltyModel=DeepSeek-R1-Distill-Qwen-7B2026.0692.22,451
Short-RLModel=DeepSeek-R1-Distill-Qwen-7B2026.0691.72,234
VanillaModel=DeepSeek-R1-Distill-Qwen-7B2026.0685.85,590
GRPOModel=DeepSeek-R1-Distill-Qwen-14B2026.06844,471