Semantic Textual Similarity On Sts15

评估指标

Spearman Correlation

评测结果

各个模型在此基准测试上的表现结果

		Paper Title
PromptEOL+CSE+LLaMA-30B	0.9004	Scaling Sentence Embeddings with Large Language Models
AnglE-LLaMA-13B	0.8956	AnglE-optimized Text Embeddings
PromptEOL+CSE+OPT-13B	0.8952	Scaling Sentence Embeddings with Large Language Models
PromptEOL+CSE+OPT-2.7B	0.8951	Scaling Sentence Embeddings with Large Language Models
AnglE-LLaMA-7B-v2	0.8943	AnglE-optimized Text Embeddings
Trans-Encoder-RoBERTa-large-cross (unsup.)	0.8863	Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
Trans-Encoder-BERT-large-bi (unsup.)	0.8816	Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
PromCSE-RoBERTa-large (0.355B)	0.8808	Improved Universal Sentence Embeddings with Prompt-based Contrastive Learning and Energy-based Learning
SimCSE-RoBERTalarge	0.8666	SimCSE: Simple Contrastive Learning of Sentence Embeddings
Trans-Encoder-RoBERTa-base-cross (unsup.)	0.8577	Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
Trans-Encoder-BERT-base-bi (unsup.)	0.8508	Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
Trans-Encoder-BERT-base-cross (unsup.)	0.8444	Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
DiffCSE-BERT-base	0.8390	DiffCSE: Difference-based Contrastive Learning for Sentence Embeddings
DiffCSE-RoBERTa-base	0.8281	DiffCSE: Difference-based Contrastive Learning for Sentence Embeddings
SRoBERTa-NLI-large	0.8185	Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
Mirror-BERT-base (unsup.)	0.814	Fast, Effective, and Self-Supervised: Transforming Masked Language Models into Universal Lexical and Sentence Encoders
Dino (STSb/)	0.8049	Generating Datasets with Pretrained Language Models
Mirror-RoBERTa-base (unsup.)	0.798	Fast, Effective, and Self-Supervised: Transforming Masked Language Models into Universal Lexical and Sentence Encoders
IS-BERT-NLI	0.7523	An Unsupervised Sentence Embedding Method by Mutual Information Maximization
BERTlarge-flow (target)	0.7492	On the Sentence Embeddings from Pre-trained Language Models

0 of 20 row(s) selected.