List your company

Mathematical Reasoning on GSM8K (test) (Pass@1, Token Count, Ratio)

95.7Pass@1SimPOShortest
86.75689.07891.493.722Jan 20, 2026

Evaluation results

MethodLinks
SimPOShortestBackbone=Qwen3-8B2026.0195.71,24761.8
OriginalBackbone=Qwen3-8B2026.0195.52,017100
BeliefPromptBackbone=Qwen3-8B2026.0195.21,10955
RELIEFBackbone=Qwen3-8B2026.019575737.5
SFTbeliefBackbone=Qwen3-8B2026.0194.91,93495.9
SFTShortestBackbone=Qwen3-8B2026.0194.42,162107.2
SimPOShortestBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0191.31,37769.8
RELIEFBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0189.81,40071
OriginalBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0188.71,973100
SFTShortestBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0188.389745.5
SFTbeliefBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0187.71,42872.4
BeliefPromptBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0187.149525.1