Sentiment Analysis On Sst 2 Binary

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
T5-11B	97.5	Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
MT-DNN-SMART	97.5	SMART: Robust and Efficient Fine-Tuning for Pre-trained Natural Language Models through Principled Regularized Optimization
T5-3B	97.4	Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
MUPPET Roberta Large	97.4	Muppet: Massive Multi-task Representations with Pre-Finetuning
StructBERTRoBERTa ensemble	97.1	StructBERT: Incorporating Language Structures into Pre-training for Deep Language Understanding	-
ALBERT	97.1	ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
XLNet (single model)	97	XLNet: Generalized Autoregressive Pretraining for Language Understanding
ELECTRA	96.9	ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators
RoBERTa-large 355M + Entailment as Few-shot Learner	96.9	Entailment as Few-Shot Learner
XLNet-Large (ensemble)	96.8	XLNet: Generalized Autoregressive Pretraining for Language Understanding
FLOATER-large	96.7	Learning to Encode Position for Transformer with Continuous Dynamical Model
MUPPET Roberta base	96.7	Muppet: Massive Multi-task Representations with Pre-Finetuning
RoBERTa (ensemble)	96.7	RoBERTa: A Robustly Optimized BERT Pretraining Approach
DeBERTa (large)	96.5	DeBERTa: Decoding-enhanced BERT with Disentangled Attention
MT-DNN-ensemble	96.5	Improving Multi-Task Deep Neural Networks via Knowledge Distillation for Natural Language Understanding
RoBERTa-large 355M (MLP quantized vector-wise, fine-tuned)	96.4	LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale
ASA + RoBERTa	96.3	Adversarial Self-Attention for Language Understanding
T5-Large 770M	96.3	Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
Snorkel MeTaL(ensemble)	96.2	Training Complex Models with Multi-Task Weak Supervision
PSQ (Chen et al., 2020)	96.2	A Statistical Framework for Low-bitwidth Training of Deep Neural Networks

0 of 88 row(s) selected.