List your company

Mathematical Reasoning on GSM8k, MATH, SVAMP, and MM

58.15GSM8k Accuracy+ RFT w/ TD
9.342822.013934.68547.3561Jan 11, 2024

Evaluation results

MethodLinks
+ RFT w/ TDModel Parameters=13B, Base Model=SFT LLM2024.0158.1513.9858.841.5843.13
+ RLMECModel Parameters=13B, Base Model=SFT LLM2024.0158.15146045.0744.31
+ RFT w/ RDModel Parameters=13B, Base Model=SFT LLM2024.0157.3914.3456.242.8142.96
+ RFT w/ GTModel Parameters=13B, Base Model=SFT LLM2024.0157.3213.7456.743.9442.93
+ CoHModel Parameters=13B, Base Model=SFT LLM2024.0157.3113.15442.341.68
SFT LLMModel Parameters=13B2024.0156.6312.7453.541.2741.04
+ SFTModel Parameters=13B, Base Model=SFT LLM2024.0155.8813.625841.2742.19
+ RFTModel Parameters=13B, Base Model=SFT LLM2024.0155.813.6254.141.6841.3
+ Vanilla PPOModel Parameters=13B, Base Model=SFT LLM2024.0153.4511.7655.143.1240.86
+ RFT w/ TDModel Parameters=7B, Base Model=SFT LLM2024.0151.8611.0449.438.1937.62
+ RFT w/ RDModel Parameters=7B, Base Model=SFT LLM2024.0151.7811.2448.740.7638.12
+ RLMECModel Parameters=7B, Base Model=SFT LLM2024.0151.1811.1649.640.9738.23
SFT LLMModel Parameters=7B2024.0151.0210.4847.838.536.95
+ PPO A2CModel Parameters=7B, Base Model=SFT LLM2024.0150.949.3846.638.536.36
+ SFTModel Parameters=7B, Base Model=SFT LLM2024.0150.3411.0447.238.436.75
+ CoHModel Parameters=7B, Base Model=SFT LLM2024.0150.1110.9448.638.537.04
+ RFT w/ GTModel Parameters=7B, Base Model=SFT LLM2024.0149.8911.2646.738.9136.69
+ RFTModel Parameters=7B, Base Model=SFT LLM2024.0149.6610.848.339.0136.94
+ Vanilla PPOModel Parameters=7B, Base Model=SFT LLM2024.0148.9710.6444.938.635.78
+ DPOModel Parameters=13B, Base Model=SFT LLM2024.0144.24.3839.732.1430.11
+ DPOModel Parameters=7B, Base Model=SFT LLM2024.0134.195.3825.832.5824.49
WizardLMModel Parameters=13B2024.0131.013.185221.3626.89
VicunaModel Parameters=13B2024.0124.14.7433.829.9823.16
LLaMA 2Model Parameters=13B2024.0121.236.5834.434.3924.15
WizardLMModel Parameters=7B2024.0114.483.3434.827.119.93
VicunaModel Parameters=7B2024.0112.24.2624.326.0816.71
LLaMA 2Model Parameters=7B2024.0111.224.829.728.4418.54