Transfer Learning on SentEval Transfer Learning Tasks (test)
90.9MRPromptEOL + Token Prepending
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| PromptEOL + Token PrependingBackbone=LLaMA2-7B, Evaluation Protocol=Logistic Regression Classifier2024.12 | 90.9 | 93.35 | 96.58 | 91.51 | 95.5 | 96 | 76.12 | 91.42 | |
| PromptEOLBackbone=LLaMA2-7B, Evaluation Protocol=Logistic Regression Classifier2024.12 | 90.63 | 92.87 | 96.32 | 91.19 | 95 | 95.4 | 75.19 | 90.94 | |
| Pretended CoT + Token PrependingBackbone=LLaMA2-7B, Evaluation Protocol=Logistic Regression Classifier2024.12 | 90.45 | 92.61 | 96.52 | 91.59 | 95.77 | 96 | 76.81 | 91.39 | |
| Knowledge + Token PrependingBackbone=LLaMA2-7B, Evaluation Protocol=Logistic Regression Classifier2024.12 | 90.39 | 93.32 | 96.31 | 91.56 | 94.51 | 97.6 | 76.06 | 91.39 | |
| Pretended CoTBackbone=LLaMA2-7B, Evaluation Protocol=Logistic Regression Classifier2024.12 | 90.1 | 92.24 | 96.32 | 91.54 | 95.11 | 94.2 | 73.97 | 90.75 | |
| KnowledgeBackbone=LLaMA2-7B, Evaluation Protocol=Logistic Regression Classifier2024.12 | 89.84 | 93.03 | 96.21 | 91.54 | 94.78 | 97.2 | 73.91 | 90.93 | |
| SimCSE-RoBERTa_largeSupervision=Supervised, Backbone=RoBERTa-large, Auxiliary Task=MLM (lambda=0.1)2021.04 | 88.45 | 92.53 | 95.19 | 90.58 | 93.3 | 93.8 | 77.74 | 90.23 | |
| SimCSE-RoBERTa_largeSupervision=Supervised, Backbone=RoBERTa-large2021.04 | 88.12 | 92.37 | 95.11 | 90.49 | 92.75 | 91.8 | 76.64 | 89.61 | |
| SimCSE-RoBERTa_baseSupervision=Supervised, Backbone=RoBERTa-base, Auxiliary Task=MLM (lambda=0.1)2021.04 | 85.08 | 91.76 | 94.02 | 89.72 | 92.31 | 91.2 | 76.52 | 88.66 | |
| SimCSE-RoBERTa_baseSupervision=Supervised, Backbone=RoBERTa-base2021.04 | 84.92 | 92 | 94.11 | 89.82 | 91.27 | 88.8 | 75.65 | 88.08 | |
| SimCSE-RoBERTa_largeSupervision=Unsupervised, Backbone=RoBERTa-large, Auxiliary Task=MLM (lambda=0.1)2021.04 | 84.66 | 88.56 | 95.43 | 87.5 | 89.46 | 95 | 72.41 | 87.57 | |
| SBERT_baseSupervision=Supervised, Backbone=BERT-base2021.04 | 83.64 | 89.43 | 94.39 | 89.86 | 88.96 | 89.6 | 76 | 87.41 | |
| DALR-RoBERTaTraining Data=wiki+flickr, Backbone=RoBERTa2025.06 | 83.57 | 88.69 | 96.44 | 91.01 | 88.96 | 86.8 | 76.74 | 87.45 | |
| ClusterNS-RoBERTabasew/ MLM=true, backbone=RoBERTa-base2023.05 | 83.51 | 88.11 | 94.56 | 86.04 | 88.85 | 92.4 | 76.7 | 87.17 | |
| SimCSE-RoBERTaTraining Data=wiki, Backbone=RoBERTa2025.06 | 83.37 | 87.76 | 95.05 | 87.16 | 89.02 | 90.8 | 75.13 | 86.9 | |
| SimCSE-RoBERTabasew/ MLM=true, backbone=RoBERTa-base2023.05 | 83.37 | 87.76 | 95.05 | 87.16 | 89.02 | 90.8 | 75.13 | 86.9 | |
| KDMCSE-RoBERTaTraining Data=wiki+flickr, Backbone=RoBERTa2025.06 | 83.21 | 88.16 | 95.73 | 90.46 | 88.05 | 86.3 | 76.18 | 86.87 | |
| DALR-BERTTraining Data=wiki+flickr, Backbone=BERT2025.06 | 82.95 | 88.1 | 95.89 | 90.83 | 88.04 | 86.6 | 76.06 | 86.92 | |
| SimCSE-BERTTraining Data=wiki, Backbone=BERT2025.06 | 82.92 | 87.23 | 95.71 | 88.73 | 86.81 | 87.01 | 78.07 | 86.64 | |
| SimCSE-BERT_baseSupervision=Unsupervised, Backbone=BERT-base, Auxiliary Task=MLM (lambda=0.1)2021.04 | 82.92 | 87.23 | 95.71 | 88.73 | 86.81 | 87.01 | 78.07 | 86.64 | |
| SimCSE-BERTbasew/ MLM=true, backbone=BERT-base2023.05 | 82.92 | 87.23 | 95.71 | 88.73 | 86.81 | 87.01 | 78.07 | 86.64 | |
| MCSE-RoBERTaTraining Data=wiki+flickr, Backbone=RoBERTa2025.06 | 82.82 | 88.04 | 95.7 | 90.13 | 87.09 | 84.97 | 75.51 | 86.29 | |
| ClusterNS-BERTbasew/ MLM=true, backbone=BERT-base2023.05 | 82.79 | 86.84 | 95.29 | 88.04 | 86.88 | 91.8 | 76.99 | 86.95 | |
| KDMCSE-BERTTraining Data=wiki+flickr, Backbone=BERT2025.06 | 82.78 | 87.89 | 95.37 | 90.08 | 87.61 | 86.08 | 75.88 | 86.53 | |
| SimCSE-RoBERTa_largeSupervision=Unsupervised, Backbone=RoBERTa-large2021.04 | 82.74 | 87.87 | 93.66 | 88.22 | 88.58 | 92 | 69.68 | 86.11 | |
| DALR-RoBERTaTraining Data=wiki+coco, Backbone=RoBERTa2025.06 | 82.71 | 88.02 | 96.1 | 90.21 | 87.85 | 86.38 | 75.84 | 86.73 | |
| SimCSE-BERT_baseSupervision=Supervised, Backbone=BERT-base2021.04 | 82.69 | 89.25 | 94.81 | 89.59 | 87.31 | 88.4 | 73.51 | 86.51 | |
| SimCSE-BERT_baseSupervision=Supervised, Backbone=BERT-base, Auxiliary Task=MLM (lambda=0.1)2021.04 | 82.68 | 88.88 | 94.52 | 89.82 | 88.41 | 87.6 | 76.12 | 86.86 | |
| DALR-BERTTraining Data=wiki+coco, Backbone=BERT2025.06 | 82.66 | 87.9 | 95.85 | 90.43 | 87.59 | 86.09 | 75.74 | 86.61 | |
| KDMCSE-RoBERTaTraining Data=wiki+coco, Backbone=RoBERTa2025.06 | 82.47 | 87.88 | 95.24 | 89.95 | 87.51 | 85.77 | 75.82 | 86.37 | |
| KDMCSE-BERTTraining Data=wiki+coco, Backbone=BERT2025.06 | 82.3 | 87.71 | 95.04 | 89.86 | 87.38 | 85.68 | 75.51 | 86.2 | |
| MCSE-RoBERTaTraining Data=wiki+coco, Backbone=RoBERTa2025.06 | 82.24 | 87.53 | 95.22 | 89.76 | 87.08 | 84.15 | 74.96 | 85.85 | |
| MCSE-BERTTraining Data=wiki+flickr, Backbone=BERT2025.06 | 82.07 | 87.28 | 94.96 | 89.61 | 86.58 | 84.04 | 74.93 | 85.64 | |
| ClusterNS-BERTbasew/ MLM=false, backbone=BERT-base2023.05 | 82.01 | 85.46 | 94.44 | 89.09 | 86.27 | 88.8 | 73.57 | 85.66 | |
| SRoBERTa_baseSupervision=Supervised, Backbone=RoBERTa-base2021.04 | 81.91 | 90.83 | 92.56 | 88.75 | 90.5 | 88.6 | 78.14 | 87.76 | |
| ClusterNS-RoBERTabasew/ MLM=false, backbone=RoBERTa-base2023.05 | 81.78 | 86.65 | 93.21 | 87.85 | 87.53 | 84 | 76.46 | 85.35 | |
| MCSE-BERTTraining Data=wiki+coco, Backbone=BERT2025.06 | 81.75 | 86.89 | 94.73 | 89.44 | 86.81 | 83.97 | 74.66 | 85.46 | |
| InferSent-GloVeSupervision=Supervised, Backbone=GloVe2021.04 | 81.57 | 86.54 | 92.5 | 90.38 | 84.18 | 88.2 | 75.77 | 85.59 | |
| SimCSE-RoBERTa_baseSupervision=Unsupervised, Backbone=RoBERTa-base, Auxiliary Task=MLM (lambda=0.1)2021.04 | 81.37 | 87.76 | 95.05 | 87.16 | 89.02 | 90.8 | 75.13 | 86.9 | |
| SimCSE-BERT_baseSupervision=Unsupervised, Backbone=BERT-base2021.04 | 81.18 | 86.46 | 94.45 | 88.88 | 85.5 | 89.8 | 74.43 | 85.81 | |
| SimCSE-BERTbasew/ MLM=false, backbone=BERT-base2023.05 | 81.18 | 86.46 | 94.45 | 88.88 | 85.5 | 89.8 | 74.43 | 85.81 | |
| IS-BERT_baseSupervision=Unsupervised, Backbone=BERT-base2021.04 | 81.09 | 87.18 | 94.96 | 88.75 | 85.96 | 88.64 | 74.24 | 85.83 | |
| SimCSE-RoBERTa_baseSupervision=Unsupervised, Backbone=RoBERTa-base2021.04 | 81.04 | 87.74 | 93.28 | 86.94 | 86.6 | 84.6 | 73.68 | 84.84 | |
| SimCSE-RoBERTabasew/ MLM=false, backbone=RoBERTa-base2023.05 | 81.04 | 87.74 | 93.28 | 86.94 | 86.6 | 84.6 | 73.68 | 84.84 | |
| Universal Sentence EncoderSupervision=Supervised2021.04 | 80.09 | 85.19 | 93.98 | 86.7 | 86.38 | 93.2 | 70.14 | 85.1 | |
| BERT-[CLS] embeddingSupervision=Unsupervised, Backbone=BERT, Strategy=[CLS]2021.04 | 78.68 | 84.85 | 94.21 | 88.23 | 84.13 | 91.4 | 71.13 | 84.66 | |
| BERT-[CLS] embedding2023.05 | 78.68 | 84.85 | 94.21 | 88.23 | 84.13 | 91.4 | 71.13 | 84.66 | |
| Avg. BERT embeddingsSupervision=Unsupervised, Backbone=BERT, Strategy=Average2021.04 | 78.66 | 86.25 | 94.37 | 88.66 | 84.4 | 92.8 | 69.54 | 84.94 | |
| Avg. BERT embeddings2023.05 | 78.66 | 86.25 | 94.37 | 88.66 | 84.4 | 92.8 | 69.54 | 84.94 | |
| GloVe embeddings (avg.)Supervision=Unsupervised, Strategy=Average2021.04 | 77.25 | 78.3 | 91.17 | 87.85 | 80.18 | 83 | 72.87 | 81.52 | |
| GloVe embeddings2023.05 | 77.25 | 78.3 | 91.17 | 87.85 | 80.18 | 83 | 72.87 | 81.52 | |
| Skip-thoughtSupervision=Unsupervised2021.04 | 76.5 | 80.1 | 93.6 | 87.1 | 82 | 92.2 | 73 | 83.5 |