Natural Language Inference On Rte

评估指标

Accuracy

评测结果

各个模型在此基准测试上的表现结果

		Paper Title	Repository
Vega v2 6B (KD-based prompt transfer)	96%	Toward Efficient Language Model Pretraining and Downstream Adaptation via Self-Evolution: A Case Study on SuperGLUE	-
PaLM 540B (fine-tuned)	95.7%	PaLM: Scaling Language Modeling with Pathways
Turing NLR v5 XXL 5.4B (fine-tuned)	94.1%	Toward Efficient Language Model Pretraining and Downstream Adaptation via Self-Evolution: A Case Study on SuperGLUE	-
ST-MoE-32B 269B (fine-tuned)	93.5%	ST-MoE: Designing Stable and Transferable Sparse Expert Models
DeBERTa-1.5B	93.2%	DeBERTa: Decoding-enhanced BERT with Disentangled Attention
MUPPET Roberta Large	92.8%	Muppet: Massive Multi-task Representations with Pre-Finetuning
DeBERTaV3large	92.7%	DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing
T5-XXL 11B	92.5%	SMART: Robust and Efficient Fine-Tuning for Pre-trained Natural Language Models through Principled Regularized Optimization
T5-XXL 11B (fine-tuned)	92.5%	Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
UL2 20B (fine-tuned)	92.1%	UL2: Unifying Language Learning Paradigms
ST-MoE-L 4.1B (fine-tuned)	92.1%	ST-MoE: Designing Stable and Transferable Sparse Expert Models
SMARTRoBERTa	92.0%	SMART: Robust and Efficient Fine-Tuning for Pre-trained Natural Language Models through Principled Regularized Optimization
FLAN 137B (prompt-tuned)	91.7%	Finetuned Language Models Are Zero-Shot Learners
T5-XL 3B	91.1%	Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
RoBERTa-large 355M + Entailment as Few-shot Learner	90.5%	Entailment as Few-Shot Learner
ALBERT	89.2%	ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
Adv-RoBERTa ensemble	88.7%	StructBERT: Incorporating Language Structures into Pre-training for Deep Language Understanding	-
RoBERTa	88.2%	RoBERTa: A Robustly Optimized BERT Pretraining Approach
RoBERTa (ensemble)	88.2%	RoBERTa: A Robustly Optimized BERT Pretraining Approach
T5-Large 738M	87.4%	LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions

0 of 90 row(s) selected.