Question Answering on ECQA, QASC, OBQA, ARC
79.49ECQA Accuracy+ RLMEC
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| + RLMECModel Parameters=13B, Base Model=SFT LLM2024.01 | 79.49 | 64.15 | 65.6 | 65.19 | 68.61 | |
| + DPOModel Parameters=13B, Base Model=SFT LLM2024.01 | 78.26 | 61.45 | 62.2 | 63.8 | 66.43 | |
| + RFT w/ TDModel Parameters=13B, Base Model=SFT LLM2024.01 | 76.71 | 61.56 | 61.8 | 64.14 | 66.05 | |
| + RFT w/ GTModel Parameters=13B, Base Model=SFT LLM2024.01 | 76.66 | 60.37 | 63.4 | 63.17 | 65.9 | |
| + RFT w/ RDModel Parameters=13B, Base Model=SFT LLM2024.01 | 76.62 | 62.2 | 63.2 | 63.17 | 66.3 | |
| + CoHModel Parameters=13B, Base Model=SFT LLM2024.01 | 76.62 | 60.37 | 59.8 | 63.93 | 65.18 | |
| + Vanilla PPOModel Parameters=13B, Base Model=SFT LLM2024.01 | 76.34 | 57.99 | 61.8 | 62.29 | 64.61 | |
| SFT LLMModel Parameters=13B2024.01 | 76.12 | 59.4 | 60.8 | 62.46 | 64.7 | |
| + SFTModel Parameters=13B, Base Model=SFT LLM2024.01 | 75.89 | 57.87 | 63.4 | 62.5 | 64.92 | |
| + RFTModel Parameters=13B, Base Model=SFT LLM2024.01 | 75.71 | 60.48 | 61 | 64.06 | 65.31 | |
| + RLMECModel Parameters=7B, Base Model=SFT LLM2024.01 | 73.66 | 59.5 | 56.8 | 58.5 | 62.12 | |
| + RFT w/ TDModel Parameters=7B, Base Model=SFT LLM2024.01 | 73.11 | 58.21 | 54.2 | 57.53 | 60.76 | |
| + Vanilla PPOModel Parameters=7B, Base Model=SFT LLM2024.01 | 72.88 | 50.22 | 43.4 | 56.27 | 55.69 | |
| + RFT w/ GTModel Parameters=7B, Base Model=SFT LLM2024.01 | 72.47 | 58.53 | 53.6 | 57.11 | 60.43 | |
| + RFT w/ RDModel Parameters=7B, Base Model=SFT LLM2024.01 | 72.47 | 59.29 | 54.6 | 57.03 | 60.85 | |
| + DPOModel Parameters=7B, Base Model=SFT LLM2024.01 | 72.47 | 58.53 | 55.4 | 55.26 | 60.42 | |
| + RFTModel Parameters=7B, Base Model=SFT LLM2024.01 | 72.24 | 58.64 | 55.2 | 57.15 | 60.81 | |
| SFT LLMModel Parameters=7B2024.01 | 71.88 | 55.4 | 52 | 56.27 | 58.89 | |
| + CoHModel Parameters=7B, Base Model=SFT LLM2024.01 | 71.06 | 54.86 | 51.4 | 56.61 | 58.48 | |
| + PPO A2CModel Parameters=7B, Base Model=SFT LLM2024.01 | 70.83 | 55.08 | 52.4 | 56.02 | 58.58 | |
| + SFTModel Parameters=7B, Base Model=SFT LLM2024.01 | 70.65 | 55.94 | 51.6 | 56.99 | 58.8 | |
| + FIGAModel Parameters=7B, Base Model=SFT LLM2024.01 | 69.83 | 52.48 | 51 | 46.21 | 54.88 | |
| LLaMA 2Model Parameters=13B2024.01 | 61.53 | 45.46 | 57.9 | 64.31 | 57.3 | |
| + FIGAModel Parameters=13B, Base Model=SFT LLM2024.01 | 61.21 | 60.26 | 52.8 | 46.34 | 55.15 | |
| LLaMA 2Model Parameters=7B2024.01 | 55.97 | 39.74 | 48.4 | 52.48 | 49.15 | |
| WizardLMModel Parameters=13B2024.01 | 52.6 | 40.93 | 52.3 | 58.96 | 51.2 | |
| VicunaModel Parameters=13B2024.01 | 50.14 | 39.96 | 48.4 | 53.7 | 48.05 | |
| VicunaModel Parameters=7B2024.01 | 49.82 | 32.18 | 46.4 | 51.52 | 44.98 | |
| WizardLMModel Parameters=7B2024.01 | 36.28 | 18.68 | 27.8 | 46.59 | 32.34 |