Commonsense Plausibility Estimation (Free-form Text) on CFC
49.01Spearman Correlation (ρ)Mistral_w/ CL
Evaluation results
| Method | Links | |
|---|---|---|
| Mistral_w/ CLInference Parameters=7B, Model Category=LM, Method Framework=ComPaSS, Contrastive Learning=Yes2025.02 | 49.01 | |
| Mistral+likelihoodInference Parameters=7B, Model Category=LM, Evaluation Method=Likelihood, Method Framework=Baseline2025.02 | 47.98 | |
| Qwen2_w/ CLInference Parameters=7B, Model Category=LM, Method Framework=ComPaSS, Contrastive Learning=Yes2025.02 | 46.41 | |
| VERA-T5Inference Parameters=5B, Model Category=CSM, Method Framework=Baseline2025.02 | 45.81 | |
| Qwen2+likelihoodInference Parameters=7B, Model Category=LM, Evaluation Method=Likelihood, Method Framework=Baseline2025.02 | 45.1 | |
| RoBERTa_w/ CLInference Parameters=355M, Model Category=LM, Method Framework=ComPaSS, Contrastive Learning=Yes2025.02 | 44.46 | |
| Qwen2_w/ CL+likelihoodInference Parameters=7B, Model Category=LM, Evaluation Method=Likelihood, Method Framework=Baseline, Contrastive Learning=Yes2025.02 | 43 | |
| RoBERTa+likelihoodInference Parameters=355M, Model Category=LM, Evaluation Method=Likelihood, Method Framework=Baseline2025.02 | 42.46 | |
| EVA-CLIPInference Parameters=695M, Model Category=VLM, Method Framework=ComPaSS2025.02 | 41.46 | |
| RoBERTa_w/ CL+likelihoodInference Parameters=355M, Model Category=LM, Evaluation Method=Likelihood, Method Framework=Baseline, Contrastive Learning=Yes2025.02 | 38.03 | |
| CLIPInference Parameters=124M, Model Category=VLM, Method Framework=ComPaSS2025.02 | 35.13 | |
| ACCENTInference Parameters=440M, Model Category=CSM, Method Framework=Baseline2025.02 | 35.04 | |
| Mistral+verbal.Inference Parameters=7B, Model Category=LM, Evaluation Method=Verbalization, Method Framework=Baseline2025.02 | 32.06 | |
| Qwen2+verbal.Inference Parameters=7B, Model Category=LM, Evaluation Method=Verbalization, Method Framework=Baseline2025.02 | 29.32 |