Scaling Sentence Embeddings with Large Language Models

Open paper
Benchmark Model Rank Results
semantic-textual-similarity-on-sickPromptEOL+CSE+LLaMA-30B#2Spearman Correlation: 0.8238
semantic-textual-similarity-on-sickPromptEOL+CSE+OPT-13B#3Spearman Correlation: 0.8206
semantic-textual-similarity-on-sickPromptEOL+CSE+OPT-2.7B#5Spearman Correlation: 0.8129
semantic-textual-similarity-on-sts-benchmarkPromptEOL+CSE+LLaMA-30B#26Spearman Correlation: 0.8914
semantic-textual-similarity-on-sts-benchmarkPromptEOL+CSE+OPT-13B#30Spearman Correlation: 0.8856
semantic-textual-similarity-on-sts-benchmarkPromptEOL+CSE+OPT-2.7B#31Spearman Correlation: 0.8833
semantic-textual-similarity-on-sts12PromptEOL+CSE+OPT-13B#1Spearman Correlation: 0.8020
semantic-textual-similarity-on-sts12PromptEOL+CSE+LLaMA-30B#2Spearman Correlation: 0.7972
semantic-textual-similarity-on-sts12PromptEOL+CSE+OPT-2.7B#4Spearman Correlation: 0.7949
semantic-textual-similarity-on-sts13PromptEOL+CSE+LLaMA-30B#3Spearman Correlation: 0.9025
semantic-textual-similarity-on-sts13PromptEOL+CSE+OPT-13B#4Spearman Correlation: 0.9024
semantic-textual-similarity-on-sts13PromptEOL+CSE+OPT-2.7B#5Spearman Correlation: 0.8964
semantic-textual-similarity-on-sts14PromptEOL+CSE+LLaMA-30B#2Spearman Correlation: 0.8585
semantic-textual-similarity-on-sts14PromptEOL+CSE+OPT-13B#5Spearman Correlation: 0.8534
semantic-textual-similarity-on-sts14PromptEOL+CSE+OPT-2.7B#6Spearman Correlation: 0.8480
semantic-textual-similarity-on-sts15PromptEOL+CSE+LLaMA-30B#1Spearman Correlation: 0.9004
semantic-textual-similarity-on-sts15PromptEOL+CSE+OPT-13B#3Spearman Correlation: 0.8952
semantic-textual-similarity-on-sts15PromptEOL+CSE+OPT-2.7B#4Spearman Correlation: 0.8951
semantic-textual-similarity-on-sts16PromptEOL+CSE+LLaMA-30B#4Spearman Correlation: 0.8627
semantic-textual-similarity-on-sts16PromptEOL+CSE+OPT-2.7B#5Spearman Correlation: 0.8591
semantic-textual-similarity-on-sts16PromptEOL+CSE+OPT-13B#6Spearman Correlation: 0.8590

Rank counts only results with a code link.