Semantic Textual Similarity On Sts Benchmark

評価指標

Spearman Correlation

評価結果

このベンチマークにおける各モデルのパフォーマンス結果

モデル名	Spearman Correlation	Paper Title	Repository
T5-Large 770M	0.886	Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
SRoBERTa-NLI-STSb-large	0.8615	Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
USE_T	-	Universal Sentence Encoder
Q8BERT (Zafrir et al., 2019)	-	Q8BERT: Quantized 8Bit BERT
DistilBERT 66M	-	DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter
RoBERTa	-	RoBERTa: A Robustly Optimized BERT Pretraining Approach
SBERT-NLI-base	0.7703	Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
StructBERTRoBERTa ensemble	0.924	StructBERT: Incorporating Language Structures into Pre-training for Deep Language Understanding	-
Dino (STSb/̄	0.7782	Generating Datasets with Pretrained Language Models
SRoBERTa-NLI-base	0.7777	Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
ERNIE 2.0 Large	-	ERNIE 2.0: A Continual Pre-training Framework for Language Understanding
ALBERT	-	ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
SMART-BERT	-	SMART: Robust and Efficient Fine-Tuning for Pre-trained Natural Language Models through Principled Regularized Optimization
AnglE-LLaMA-13B	0.8969	AnglE-optimized Text Embeddings
Trans-Encoder-RoBERTa-large-cross (unsup.)	0.867	Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
ERNIE	-	ERNIE: Enhanced Language Representation with Informative Entities
SBERT-NLI-large	0.79	Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
AnglE-LLaMA-7B	0.8897	AnglE-optimized Text Embeddings
Trans-Encoder-BERT-base-bi (unsup.)	0.839	Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
Mirror-BERT-base (unsup.)	0.764	Fast, Effective, and Self-Supervised: Transforming Masked Language Models into Universal Lexical and Sentence Encoders

0 of 66 row(s) selected.