Semantic Textual Similarity on STS Benchmark (dev)
0.926Pearson Correlation (r)UNIMO
Evaluation results
| Method | Links | |
|---|---|---|
| UNIMOSize=large2020.12 | 0.926 | |
| RoBERTaSize=large2020.12 | 0.924 | |
| XLNetSize=large2020.12 | 0.918 | |
| UNIMOSize=base2020.12 | 0.91 | |
| BERTSize=large2020.12 | 0.9 | |
| UNIMO (w/o single-modal)Size=base, Note=Removing single-modal learning process2020.12 | 0.888 | |
| UniLMSize=large2020.12 | 0.877 | |
| ECNU2018.04 | 0.847 | |
| Reddit+SNLIFine-tuned on STS=true2018.04 | 0.835 | |
| CNN (HCTI)2018.04 | 0.834 | |
| BIT2018.04 | 0.829 | |
| Reddit+SNLIFine-tuned on STS=false2018.04 | 0.814 | |
| USE TUniversal Encoder=Transformer (USE_T), Transfer Level=Sentence, Embedding Initialization=None, Transfer Task Model=None2018.03 | 0.814 | |
| RedditFine-tuned on STS=true2018.04 | 0.809 | |
| InferSent2018.04 | 0.801 | |
| SIF2018.04 | 0.801 | |
| Sent2Vec2018.04 | 0.787 | |
| USE DUniversal Encoder=Deep Averaging Network (USE_D), Transfer Level=Sentence, Embedding Initialization=None, Transfer Task Model=None2018.03 | 0.763 | |
| RedditFine-tuned on STS=false2018.04 | 0.762 | |
| C-PHRASE2018.04 | 0.743 | |
| PV-DBOW2018.04 | 0.722 |