Running the STS benchmark for the SimCSE model and the Fast Sentence Embeddings model.
The benchmark is the STS benchmark and scripts from the benchmark to run Pearson correlation, have been used here.
Files have been downloaded from the STS benchmark site. We only take the sts-dev file to run the benchmark.
SimCSE scores around 65.98 while FSE scores around 45.5 for glove-wiki-gigaword-50 and around 63.135 for paranmt-300 and using the Averages model.