Semantic Textual Similarity on STS tasks (STS12, STS13, STS14, STS15, STS16, STS-B, SICK-R)
80.67STS12 ScoreModernBERT_emb
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| ModernBERT_embDimension=1,024, Type=Dense2025.05 | 80.67 | 87.87 | 83.8 | 88.59 | 86.82 | 87.4 | 80.31 | 85.07 | |
| PromCSE-RoBERTalarge + EHlearning_mode=supervised, backbone=RoBERTa-large, loss_enhancement=Energy-based Hinge loss (EH)2022.03 | 79.56 | 88.97 | 83.81 | 88.08 | 84.96 | 87.87 | 82.43 | 85.1 | |
| LDIRBackbone=AngIE, Dimension=200, Type=Dense2025.05 | 79.28 | 82.16 | 80.96 | 84.64 | 83.68 | 85.32 | 77.92 | 81.99 | |
| AngIEDimension=1,024, Type=Dense2025.05 | 79.09 | 89.62 | 85.02 | 89.51 | 86.61 | 89.06 | 82.62 | 85.93 | |
| LDIRBackbone=AngIE, Dimension=500, Type=Dense2025.05 | 78.85 | 84.35 | 80.93 | 84.79 | 83.61 | 86.31 | 80.85 | 82.82 | |
| DYSEM (Language-Specific Prompt)Backbone=Qwen3-8B, #Dim.=675*2026.05 | 77.54 | 84.33 | 76.95 | 83.91 | 81.19 | 81.26 | 78.56 | 80.54 | |
| DYSEMBase Model=Qwen2.5-7B-it, #Dim.=1001*, Prompt Type=English Prompt2026.05 | 77.54 | 82.28 | 75.35 | 83.49 | 78.89 | 80.49 | 78.08 | 79.44 | |
| PromCSE-RoBERTabase + EHlearning_mode=supervised, backbone=RoBERTa-base, loss_enhancement=Energy-based Hinge loss (EH)2022.03 | 77.51 | 86.15 | 81.59 | 86.92 | 83.81 | 86.35 | 80.49 | 83.26 | |
| SimCSE-RoBERTa_largesupervision=supervised2021.04 | 77.46 | 87.27 | 82.36 | 86.66 | 83.93 | 86.7 | 81.95 | 83.76 | |
| DYSEM (English Prompt)Backbone=Qwen2.5-7B, #Dim.=978*2026.05 | 76.92 | 83.2 | 75.48 | 83.51 | 79.57 | 80.29 | 78.43 | 79.63 | |
| SimCSE-RoBERTa_basesupervision=supervised2021.04 | 76.53 | 85.21 | 80.95 | 86.03 | 82.57 | 85.83 | 80.5 | 82.52 | |
| SimCSE-RoBERTabaselearning_mode=supervised, backbone=RoBERTa-base2022.03 | 76.53 | 85.21 | 80.95 | 86.03 | 82.57 | 85.83 | 80.5 | 82.52 | |
| SimCSEBackbone=RoBERTa-base2023.05 | 76.53 | 85.21 | 80.95 | 86.03 | 82.57 | 85.83 | 80.5 | 82.52 | |
| DYSEMBase Model=Qwen3-8B-it, #Dim.=728*, Prompt Type=Language-Specific Prompt2026.05 | 76.43 | 84.13 | 76.29 | 83.81 | 81.2 | 81.75 | 77.5 | 80.16 | |
| PromptRoBERTaBackbone=RoBERTa-base2023.05 | 76.41 | 85.64 | 82.11 | 86.18 | 82.71 | 85.74 | 79.95 | 82.68 | |
| DYSEMBase Model=LLaMA3.1-8B-it, #Dim.=772*, Prompt Type=Language-Specific Prompt2026.05 | 76.38 | 84.87 | 77.23 | 83.92 | 82.5 | 83.77 | 79.74 | 81.2 | |
| PromptRoBERTa w/ CLHAIFBackbone=RoBERTa-base2023.05 | 76.26 | 86.01 | 82.83 | 86.7 | 82.94 | 86.04 | 80.55 | 83.05 | |
| SimCSE w/ CLHAIFBackbone=RoBERTa-base2023.05 | 76.23 | 85.46 | 81.48 | 86.47 | 83.4 | 85.93 | 80.95 | 82.85 | |
| PromCSE-BERTbase + EHlearning_mode=supervised, backbone=BERT-base, loss_enhancement=Energy-based Hinge loss (EH)2022.03 | 75.96 | 84.99 | 80.44 | 86.83 | 81.3 | 81.1 | 80.96 | 82.13 | |
| Mirror-RoBERTa-base-drophead + TENCBackbone=RoBERTa-base, Head Configuration=drophead, Pre-training Method=Mirror-BERT, TENC Configuration=cross2021.09 | 75.7 | 84.58 | 78.35 | 86.49 | 83.96 | 84.26 | 72.76 | 80.87 | |
| PromCSE-BERTbaselearning_mode=supervised, backbone=BERT-base2022.03 | 75.58 | 84.33 | 79.67 | 85.79 | 81.24 | 84.25 | 80.79 | 81.81 | |
| SimCSE-BERT_basesupervision=supervised2021.04 | 75.3 | 84.67 | 80.19 | 85.4 | 80.82 | 84.25 | 80.39 | 81.57 | |
| SimCSEBackbone=BERT-base2023.05 | 75.3 | 84.67 | 80.19 | 85.4 | 80.82 | 84.25 | 80.39 | 81.57 | |
| SimCSE_supDimension=768, Type=Dense2025.05 | 75.3 | 84.67 | 80.19 | 85.4 | 80.82 | 84.25 | 68.38 | 79.86 | |
| SimCSE-BERTbase + EHlearning_mode=supervised, backbone=BERT-base, loss_enhancement=Energy-based Hinge loss (EH)2022.03 | 75.22 | 84.93 | 81.37 | 85.94 | 80.94 | 84.78 | 80.38 | 81.94 | |
| DYSEM (English Prompt)Backbone=Qwen3-8B, #Dim.=1027*2026.05 | 75.22 | 83.22 | 76.82 | 83.49 | 79.54 | 80.09 | 79.16 | 79.65 | |
| SimCSE-BERTbase (reproduce)learning_mode=supervised, backbone=BERT-base2022.03 | 75.13 | 84.35 | 80.26 | 85.45 | 80.83 | 84.29 | 80.39 | 81.53 | |
| DYSEM (Language-Specific Prompt)Backbone=Qwen2.5-7B, #Dim.=940*2026.05 | 75.12 | 82.66 | 73.94 | 81.74 | 79.8 | 78.42 | 77.94 | 78.52 | |
| DYSEMBase Model=Qwen2.5-7B-it, #Dim.=900*, Prompt Type=Language-Specific Prompt2026.05 | 75.12 | 82.03 | 74.55 | 82.2 | 79.05 | 79.82 | 78.99 | 78.82 | |
| PromptBERTBackbone=BERT-base2023.05 | 75.1 | 85.54 | 80.58 | 86 | 81.24 | 84.57 | 80.36 | 81.91 | |
| PromptBERT w/ CLHAIFBackbone=BERT-base2023.05 | 75.03 | 85.88 | 81.48 | 86.33 | 81.4 | 84.93 | 80.98 | 82.29 | |
| Mirror-RoBERTa-base-drophead + TENCBackbone=RoBERTa-base, Head Configuration=drophead, Pre-training Method=Mirror-BERT, TENC Configuration=bi2021.09 | 74.95 | 83.86 | 77.5 | 85.8 | 83.22 | 83.94 | 72.56 | 80.26 | |
| DYSEM (Language-specific Prompt)Backbone=LLaMA3.1-8B, #Dim.=995*2026.05 | 74.95 | 84.18 | 76.22 | 83.02 | 81.3 | 81.79 | 77.25 | 79.82 | |
| SimCSE w/ CLHAIFBackbone=BERT-base2023.05 | 74.86 | 85.09 | 81.24 | 85.96 | 81.33 | 84.96 | 81.36 | 82.08 | |
| CT-SBERT_basesupervision=supervised2021.04 | 74.84 | 83.2 | 78.07 | 83.84 | 77.93 | 81.46 | 76.42 | 79.39 | |
| WhitenedCSEDimension=768, Type=Dense2025.05 | 74.65 | 85.79 | 77.49 | 84.71 | 80.33 | 81.48 | 75.34 | 79.97 | |
| S-ROBERTa-largeevaluation_mode=Zero-shot, structure=Siamese, fine-tuning_data=SNLI + MNLI2020.04 | 74.53 | 77 | 73.18 | 81.85 | 76.82 | 79.1 | 74.29 | 76.68 | |
| SBERT_oriDimension=768, Type=Dense2025.05 | 74.53 | 77 | 73.18 | 81.85 | 76.82 | 79.1 | 74.29 | 76.68 | |
| S-SegaBERT-largeevaluation_mode=Zero-shot, structure=Siamese, fine-tuning_data=SNLI + MNLI2020.04 | 74.49 | 78.64 | 74.88 | 83.28 | 77.1 | 79.42 | 73.77 | 77.37 | |
| LDIRBackbone=SBERT_new, Dimension=200, Type=Dense2025.05 | 74.44 | 77.73 | 73.58 | 81.93 | 77.98 | 80.82 | 78.42 | 77.84 | |
| DYSEMBase Model=LLaMA3.1-8B-it, #Dim.=1176*, Prompt Type=English Prompt2026.05 | 74.41 | 83.62 | 76.11 | 82.77 | 81.01 | 82.56 | 77.84 | 79.76 | |
| KDMCSE-BERTTraining Data=wiki+flickr, Backbone=BERT2025.06 | 74.4 | 83.1 | 76.3 | 83.7 | 78.8 | 81.3 | 73 | 78.6 | |
| CLAIF_scaledBackbone=BERT-base2023.05 | 74.36 | 85.07 | 80.64 | 87.21 | 83.36 | 86.26 | 79.68 | 82.37 | |
| ConSERTBackbone=BERT-base2023.05 | 74.07 | 83.93 | 77.05 | 83.66 | 78.76 | 81.36 | 76.77 | 79.37 | |
| DYSEMBase Model=Qwen3-8B-it, #Dim.=1066*, Prompt Type=English Prompt2026.05 | 74.04 | 83.18 | 76.03 | 82.8 | 79.31 | 79.93 | 76.44 | 78.82 | |
| PromptROBERTaBackbone=RoBERTa-base2023.05 | 73.94 | 84.74 | 77.28 | 84.99 | 81.74 | 81.88 | 69.5 | 79.15 | |
| DALR-BERTTraining Data=wiki+flickr, Backbone=BERT2025.06 | 73.9 | 84 | 76.5 | 84.3 | 80.6 | 81.8 | 75.3 | 79.5 | |
| KDMCSE-RoBERTaTraining Data=wiki+flickr, Backbone=RoBERTa2025.06 | 73.6 | 83.8 | 77.4 | 84 | 81.5 | 82.3 | 71.2 | 79.1 | |
| DALR-RoBERTaTraining Data=wiki+flickr, Backbone=RoBERTa2025.06 | 73.6 | 84.4 | 77.2 | 84.9 | 82 | 82.6 | 74.6 | 79.9 | |
| DALR-BERTTraining Data=wiki+coco, Backbone=BERT2025.06 | 73.4 | 82.6 | 75.6 | 83.5 | 80.8 | 80.5 | 74.1 | 78.6 | |
| KDMCSE-BERTTraining Data=wiki+coco, Backbone=BERT2025.06 | 73.2 | 80.5 | 75.4 | 83.2 | 79.7 | 79.7 | 73.7 | 77.9 | |
| DALR-RoBERTaTraining Data=wiki+coco, Backbone=RoBERTa2025.06 | 73.1 | 83.2 | 76.5 | 83.9 | 82.2 | 81.2 | 72 | 78.9 | |
| Mirror-RoBERTa-base + TENCBackbone=RoBERTa-base, Pre-training Method=Mirror-BERT, TENC Configuration=cross2021.09 | 73.1 | 82.31 | 76.49 | 83.71 | 79.64 | 81.7 | 69.7 | 78.09 | |
| SBERT_newDimension=768, Type=Dense2025.05 | 73.08 | 82.13 | 76.73 | 85.58 | 80.23 | 83.09 | 79.32 | 80.02 | |
| PromCSE-BERTbaselearning_mode=unsupervised2022.03 | 73.03 | 85.18 | 76.7 | 84.19 | 79.69 | 80.62 | 70 | 78.49 | |
| LDIRBackbone=ModernBERT_emb, Dimension=200, Type=Dense2025.05 | 72.99 | 79.63 | 75.09 | 82.52 | 79.93 | 82.61 | 78.43 | 78.74 | |
| SimCSE-RoBERTa_largesupervision=unsupervised2021.04 | 72.86 | 83.99 | 75.62 | 84.77 | 81.8 | 81.98 | 71.26 | 78.9 | |
| ROBERTa-LargeTraining=SimCSE2025.01 | 72.86 | 83.99 | 75.62 | 84.77 | 81.8 | 81.98 | 71.26 | 78.9 | |
| text-embedding-ada-002Dimension=1,536, Type=Dense2025.05 | 72.84 | 86.1 | 81.15 | 88.49 | 85.08 | 83.56 | 79 | 82.31 | |
| KDMCSE-RoBERTaTraining Data=wiki+coco, Backbone=RoBERTa2025.06 | 72.8 | 81.7 | 76.1 | 83.4 | 81.5 | 80.7 | 69.9 | 78 | |
| DYSEM (Language-specific Prompt)Backbone=LLaMA2-7B, #Dim.=1240*2026.05 | 72.68 | 84.75 | 77.28 | 83.47 | 79.19 | 80.06 | 75.39 | 78.98 | |
| DinoGPT-3Backbone=BERT-base2023.05 | 72.61 | 81.92 | 75.09 | 80.42 | 76.26 | 77.1 | 70.43 | 76.26 | |
| CLAIF_scaledBackbone=RoBERTa-base2023.05 | 72.58 | 84.5 | 79.48 | 86.92 | 84.19 | 85.85 | 79.64 | 81.88 | |
| LDIRBackbone=SBERT_new, Dimension=500, Type=Dense2025.05 | 72.51 | 79.92 | 74.93 | 82.93 | 78.44 | 81.88 | 79.34 | 78.55 | |
| S-BERT-large*evaluation_mode=Zero-shot, structure=Siamese, fine-tuning_data=SNLI + MNLI, re-implemented=true2020.04 | 72.39 | 78.06 | 75.26 | 81.79 | 76.35 | 78.64 | 73.85 | 76.62 | |
| DYSEMBase Model=Phi3-mini-it, #Dim.=869*, Prompt Type=Language-Specific Prompt2026.05 | 72.39 | 82.41 | 74.02 | 82.81 | 79.98 | 81.2 | 76.77 | 78.51 | |
| LDIRBackbone=ModernBERT_emb, Dimension=500, Type=Dense2025.05 | 72.3 | 81.97 | 75.48 | 82.86 | 79.88 | 83.22 | 80.72 | 79.49 | |
| DiffCSE-BERTbaselearning_mode=unsupervised2022.03 | 72.28 | 84.43 | 76.47 | 83.9 | 80.54 | 80.59 | 71.23 | 78.49 | |
| DiffCSEBackbone=BERT-base2023.05 | 72.28 | 84.43 | 76.47 | 83.9 | 80.54 | 80.59 | 71.23 | 78.49 | |
| S-BERT-largeevaluation_mode=Zero-shot, structure=Siamese, fine-tuning_data=SNLI + MNLI2020.04 | 72.27 | 78.46 | 74.9 | 80.99 | 76.25 | 79.23 | 73.75 | 76.55 | |
| Mirror-RoBERTa-base + TENCBackbone=RoBERTa-base, Pre-training Method=Mirror-BERT, TENC Configuration=bi2021.09 | 71.99 | 81.85 | 75.73 | 83.32 | 79.97 | 81.19 | 69.47 | 77.64 | |
| SSCL-SimCSE_largescale=Large, batch size=64, max sequence length=322023.05 | 71.98 | 85.74 | 77.94 | 85.94 | 80.08 | 81.2 | 74.28 | 79.69 | |
| DYSEMBase Model=Phi3.5-mini-it, #Dim.=997*, Prompt Type=English Prompt2026.05 | 71.87 | 81.4 | 73.49 | 80.48 | 77.27 | 78.1 | 77.36 | 77.14 | |
| MCSE-RoBERTaTraining Data=wiki+flickr, Backbone=RoBERTa2025.06 | 71.7 | 82.7 | 75.9 | 84 | 81.3 | 82.3 | 70.3 | 78.3 | |
| SSCL-SimCSEscale=Base2023.05 | 71.68 | 83.5 | 76.42 | 83.46 | 78.39 | 79.03 | 71.76 | 77.9 | |
| DYSEMBase Model=Phi3-mini-it, #Dim.=987*, Prompt Type=English Prompt2026.05 | 71.63 | 82.2 | 74 | 81.3 | 76.37 | 77.15 | 75.22 | 76.84 | |
| PromptBERTbaselearning_mode=unsupervised2022.03 | 71.56 | 84.58 | 76.98 | 84.47 | 80.6 | 81.6 | 69.87 | 78.54 | |
| PromptBERTBackbone=BERT-base2023.05 | 71.56 | 84.58 | 76.98 | 84.47 | 80.6 | 81.6 | 69.87 | 78.54 | |
| PromptBERT (110M)#Dim.=7682026.05 | 71.56 | 84.58 | 76.98 | 84.47 | 80.6 | 81.6 | 69.87 | 78.54 | |
| AlignedWVABackbone=Qwen2.5-7B, #Dim.=35842026.05 | 71.56 | 80.19 | 69.61 | 75.63 | 78.69 | 74.45 | 73.06 | 74.74 | |
| SRoBERTa_basesupervision=supervised2021.04 | 71.54 | 72.49 | 70.8 | 78.74 | 73.69 | 77.77 | 74.46 | 74.21 | |
| SRoBERTaBackbone=RoBERTa-base2023.05 | 71.54 | 72.49 | 70.8 | 78.74 | 73.69 | 77.77 | 74.46 | 74.21 | |
| MCSE-BERTTraining Data=wiki+flickr, Backbone=BERT2025.06 | 71.4 | 81.8 | 74.8 | 83.6 | 77.5 | 79.5 | 72.6 | 77.3 | |
| DYSEMBase Model=Phi3.5-mini-it, #Dim.=658*, Prompt Type=Language-Specific Prompt2026.05 | 71.36 | 81.41 | 73.65 | 81.21 | 74.65 | 78.67 | 75.76 | 76.67 | |
| DinoGPT-3Backbone=RoBERTa-base2023.05 | 71.24 | 81.55 | 75.67 | 81.42 | 78.77 | 80.1 | 71.31 | 77.15 | |
| MCSE-BERTTraining Data=wiki+coco, Backbone=BERT2025.06 | 71.2 | 79.7 | 73.8 | 83 | 77.8 | 78.5 | 72.1 | 76.6 | |
| SBERT_basesupervision=supervised2021.04 | 70.97 | 76.53 | 73.19 | 79.09 | 74.3 | 77.03 | 72.91 | 74.89 | |
| SBERTbaselearning_mode=supervised2022.03 | 70.97 | 76.53 | 73.19 | 79.09 | 74.3 | 77.03 | 72.91 | 74.89 | |
| SBERTBackbone=BERT-base2023.05 | 70.97 | 76.53 | 73.19 | 79.09 | 74.3 | 77.03 | 72.91 | 74.89 | |
| Sentence-BERT (110M)#Dim.=7682026.05 | 70.97 | 76.53 | 73.19 | 79.09 | 74.3 | 77.03 | 72.91 | 74.89 | |
| DYSEMBase Model=LLaMA2-7B-it, #Dim.=1193*, Prompt Type=Language-Specific Prompt2026.05 | 70.84 | 80.52 | 74.85 | 82.69 | 74.61 | 79.23 | 77.39 | 77.16 | |
| ConSERT_largeBackbone=BERT-large, Unsupervised Data Source=STS unlabeled texts2021.05 | 70.69 | 82.96 | 74.13 | 82.78 | 76.66 | 77.53 | 70.37 | 76.45 | |
| Consert_largescale=Large, batch size=64, max sequence length=322023.05 | 70.69 | 82.96 | 74.13 | 82.78 | 76.66 | 77.53 | 70.47 | 76.45 | |
| Contrastive-Tension-BERT-base + TENCBackbone=BERT-base, Pre-training Method=Contrastive Tension, TENC Configuration=cross2021.09 | 70.67 | 79.27 | 69.26 | 76.77 | 78.78 | 79.08 | 70.66 | 74.93 | |
| CLAIFBackbone=BERT-base2023.05 | 70.62 | 81.51 | 76.29 | 85.05 | 81.36 | 84.34 | 78.22 | 79.63 | |
| SRoBERTa_base-whiteningpost-processing=whitening, supervision=supervised2021.04 | 70.46 | 77.07 | 74.46 | 81.64 | 76.43 | 79.49 | 76.65 | 76.6 | |
| SRoBERTa-whiteningBackbone=RoBERTa-base2023.05 | 70.46 | 77.07 | 74.46 | 81.64 | 76.43 | 79.49 | 76.65 | 76.6 | |
| DinoBackbone=RoBERTa-base2023.05 | 70.27 | 81.26 | 71.25 | 80.49 | 77.18 | 77.82 | 68.09 | 75.2 | |
| MCSE-RoBERTaTraining Data=wiki+coco, Backbone=RoBERTa2025.06 | 70.2 | 82 | 75.5 | 83 | 81.5 | 80.8 | 69.9 | 77.6 | |
| DYSEM (English Prompt)Backbone=LLaMA2-7B, #Dim.=1238*2026.05 | 70.2 | 85.09 | 77.39 | 82.64 | 79.73 | 80.67 | 74.75 | 78.64 |