Semantic Textual Similarity On Sts14

평가 지표

Spearman Correlation

평가 결과

이 벤치마크에서 각 모델의 성능 결과

모델 이름	Spearman Correlation	Paper Title	Repository
DiffCSE-BERT-base	0.7647	DiffCSE: Difference-based Contrastive Learning for Sentence Embeddings
AnglE-LLaMA-7B-v2	0.8579	AnglE-optimized Text Embeddings
SimCSE-RoBERTalarge	0.8236	SimCSE: Simple Contrastive Learning of Sentence Embeddings
BERTlarge-flow (target)	0.6942	On the Sentence Embeddings from Pre-trained Language Models	-
PromptEOL+CSE+OPT-13B	0.8534	Scaling Sentence Embeddings with Large Language Models
IS-BERT-NLI	0.6121	An Unsupervised Sentence Embedding Method by Mutual Information Maximization
Trans-Encoder-BERT-large-bi (unsup.)	0.8137	Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
AnglE-LLaMA-13B	0.8689	AnglE-optimized Text Embeddings
Dino (STSb/̄	0.7125	Generating Datasets with Pretrained Language Models
Mirror-BERT-base (unsup.)	0.713	Fast, Effective, and Self-Supervised: Transforming Masked Language Models into Universal Lexical and Sentence Encoders
Mirror-RoBERTa-base (unsup.)	0.732	Fast, Effective, and Self-Supervised: Transforming Masked Language Models into Universal Lexical and Sentence Encoders
Trans-Encoder-RoBERTa-large-bi (unsup.)	0.8176	Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
DiffCSE-RoBERTa-base	0.7549	DiffCSE: Difference-based Contrastive Learning for Sentence Embeddings
Trans-Encoder-RoBERTa-large-cross (unsup.)	0.8194	Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
SBERT-NLI-large	0.7490000000000001	Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
Trans-Encoder-RoBERTa-base-cross (unsup.)	0.7903	Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
PromCSE-RoBERTa-large (0.355B)	0.8381	Improved Universal Sentence Embeddings with Prompt-based Contrastive Learning and Energy-based Learning
Trans-Encoder-BERT-base-bi (unsup.)	0.779	Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
PromptEOL+CSE+OPT-2.7B	0.8480	Scaling Sentence Embeddings with Large Language Models
AnglE-LLaMA-7B	0.8549	AnglE-optimized Text Embeddings

0 of 21 row(s) selected.