List your company

Mathematical Reasoning on SVAMP out-of-domain (test)

97AccuracyGPT-4
14.73636.09357.4578.807Oct 5, 2023Feb 23, 2024Jul 13, 2024Dec 1, 2024Apr 21, 2025Sep 9, 2025Jan 29, 2026

Evaluation results

MethodLinks
GPT-42023.1097
CausalRMEvaluation Type=Downstream RLHF2026.0193.9
GoalRMEvaluation Type=Downstream RLHF2026.0192.9
Standard RMEvaluation Type=Downstream RLHF2026.0192.1
SFTEvaluation Type=Downstream RLHF2026.0188.6
MathCoder-LBase=Llama-2, Size=70B2023.1084.9
MathCoder-CLBase=CodeLlama, Size=34B2023.1082.5
MathCoder-CLBase=CodeLlama, Size=13B2023.1078
MathCoder-LBase=Llama-2, Size=13B2023.1076.9
SIKeD (All)Model=Gemma 7B2024.1076.4
SIKeD (Adaptive)Model=Qwen 1.5B2024.1075.4
PoTModel=Gemma 7B2024.1075
PoTModel=Qwen 1.5B2024.1074.3
SIKeD (All)Model=Qwen 1.5B2024.1073.5
CombinedModel=Qwen 1.5B2024.1073.2
SIKeD (Adaptive)Model=Gemma 7B2024.1072.9
WizardMathBase=Llama-2, Size=70B2023.1071.8
MathCoder-LBase=Llama-2, Size=7B2023.1071.5
MathCoder-CLBase=CodeLlama, Size=7B2023.1070.7
CombinedModel=Gemma 7B2024.1069.4
CoTModel=Gemma 7B2024.1066.8
CoTModel=Qwen 1.5B2024.1066.3
L2MModel=Gemma 7B2024.1064.8
L2MModel=Qwen 1.5B2024.1062.3
SIKeD (Adaptive)Model=Gemma 2B2024.1058.6
PoTModel=Gemma 2B2024.1056.8
SIKeD (All)Model=Gemma 2B2024.1056.7
CombinedModel=Gemma 2B2024.1056.2
Llama-1 RFTBase=Llama-1, Size=34B2023.1055.4
InfoRMEvaluation Type=Downstream RLHF2026.0153.9
SIKeD (All)Model=Qwen 0.5B2024.1052.1
WizardMathBase=Llama-2, Size=13B2023.1051.9
SIKeD (Adaptive)Model=Qwen 0.5B2024.1051.7
PoTModel=Qwen 0.5B2024.1051.4
CombinedModel=Qwen 0.5B2024.1051.4
Llama-1 RFTBase=Llama-1, Size=13B2023.1046.5
L2MModel=Qwen 0.5B2024.1044.6
CoTModel=Qwen 0.5B2024.1043
CoTModel=Gemma 2B2024.1041.9
GalacticaSize=30B2023.1041.6
SIKeD (All)Model=SmolLM 1.7B2024.1039.5
SIKeD (Adaptive)Model=SmolLM 1.7B2024.1037.8
WizardMathBase=Llama-2, Size=7B2023.1036.1
CombinedModel=SmolLM 1.7B2024.1035.9
PoTModel=SmolLM 1.7B2024.1034.5
L2MModel=Gemma 2B2024.1031.6
GalacticaSize=6.7B2023.1025.6
CoTModel=SmolLM 1.7B2024.1022.6
Llama-1 RFTBase=Llama-1, Size=7B2023.1021.1
L2MModel=SmolLM 1.7B2024.1017.9