Natural Language Inference on SNLI

Metric: % Train Accuracy (higher is better)

LeaderboardDataset

Loading chart...

Results

Sort:

#	Model↕	% Train Accuracy▼	Extra Data	Paper	Date↕	Code
1	+ Unigram and bigram features	99.7	No	A large annotated corpus for learning natural la...	2015-08-21	Code
2	Ntumpha	99.1	No	Multi-Task Deep Neural Networks for Natural Lang...	2019-01-31	Code
3	1024D GRU encoders w/ unsupervised 'skip-thoughts' pre-training	98.8	No	Order-Embeddings of Images and Language	2015-11-19	Code
4	MT-DNN	97.2	No	Multi-Task Deep Neural Networks for Natural Lang...	2019-01-31	Code
5	Fine-Tuned LM-Pretrained Transformer	96.6	No	-	-	Code
6	300D DMAN Ensemble	96.1	No	Discourse Marker Augmented Network with Reinforc...	2019-07-23	Code
7	300D DMAN Ensemble	96.1	No	Discourse Marker Augmented Network with Reinforc...	2019-07-23	Code
8	SJRC (BERT-Large +SRL)	95.7	No	Explicit Contextual Semantics for Text Comprehen...	2018-09-08	-
9	150D Multiway Attention Network Ensemble	95.5	No	-	-	Code
10	300D DMAN	95.4	No	Discourse Marker Augmented Network with Reinforc...	2019-07-23	Code
11	300D DMAN	95.4	No	Discourse Marker Augmented Network with Reinforc...	2019-07-23	Code
12	Densely-Connected Recurrent and Co-Attentive Network Ensemble	95	No	Semantic Sentence Matching with Densely-connecte...	2018-05-29	-
13	600D BiLSTM with generalized pooling	94.9	No	Enhancing Sentence Embedding with Generalized Po...	2018-06-26	Code
14	450D DR-BiLSTM Ensemble	94.8	No	DR-BiLSTM: Dependent Reading Bidirectional LSTM ...	2018-02-15	-
15	150D Multiway Attention Network	94.5	No	-	-	Code
16	SemBERT	94.4	No	Semantics-aware BERT for Language Understanding	2019-09-05	Code
17	KIM	94.1	No	Neural Natural Language Inference Models Enhance...	2017-11-12	Code
18	450D DR-BiLSTM	94.1	No	DR-BiLSTM: Dependent Reading Bidirectional LSTM ...	2018-02-15	-
19	RE2	94	No	Simple and Effective Text Matching with Richer A...	2019-08-01	Code
20	KIM Ensemble	93.6	No	Neural Natural Language Inference Models Enhance...	2017-11-12	Code
21	600D ESIM + 300D Syntactic TreeLSTM	93.5	No	Enhanced LSTM for Natural Language Inference	2016-09-20	Code
22	Stochastic Answer Network	93.3	No	Stochastic Answer Networks for Natural Language ...	2018-04-21	Code
23	BiMPM Ensemble	93.2	No	Bilateral Multi-Perspective Matching for Natural...	2017-02-13	Code
24	MFAE	93.18	No	-	-	Code
25	Densely-Connected Recurrent and Co-Attentive Network	93.1	No	Semantic Sentence Matching with Densely-connecte...	2018-05-29	-
26	600D Gumbel TreeLSTM encoders	93.1	No	Learning to Compose Task-Specific Tree Structures	2017-07-10	Code
27	CA-MTL	92.6	No	Conditionally Adaptive Multi-Task Learning: Impr...	2020-09-19	Code
28	DEIM	92.6	No	DEIM: An effective deep encoding and interaction...	2022-03-20	-
29	300D Reinforced Self-Attention Network	92.6	No	Reinforced Self-Attention Network: a Hybrid of H...	2018-01-31	Code
30	300D CAFE Ensemble	92.5	No	Compare, Compress and Propagate: Enhancing Neura...	2017-12-30	-
31	448D Densely Interactive Inference Network (DIIN, code) Ensemble	92.3	No	Natural Language Inference over Interaction Space	2017-09-13	Code
32	ESIM + ELMo Ensemble	92.1	No	Deep contextualized word representations	2018-02-15	Code
33	300D mLSTM word-by-word attention model	92	No	Learning Natural Language Inference with LSTM	2015-12-30	Code
34	ESIM + ELMo	91.6	No	Deep contextualized word representations	2018-02-15	Code
35	512D Dynamic Meta-Embeddings	91.6	No	Dynamic Meta-Embeddings for Improved Sentence Re...	2018-04-21	Code
36	Densely-Connected Recurrent and Co-Attentive Network (encoder)	91.4	No	Semantic Sentence Matching with Densely-connecte...	2018-05-29	-
37	448D Densely Interactive Inference Network (DIIN, code)	91.2	No	Natural Language Inference over Interaction Space	2017-09-13	Code
38	300D Gumbel TreeLSTM encoders	91.2	No	Learning to Compose Task-Specific Tree Structures	2017-07-10	Code
39	300D Directional self-attention network encoders	91.1	No	DiSAN: Directional Self-Attention Network for RN...	2017-09-14	Code
40	600D Residual stacked encoders	91	No	Shortcut-Stacked Sentence Encoders for Multi-Dom...	2017-08-07	Code
41	BiMPM	90.9	No	Bilateral Multi-Perspective Matching for Natural...	2017-02-13	Code
42	300D re-read LSTM	90.7	No	-	-	-
43	300D re-read LSTM	90.7	No	-	-	-
44	200D decomposable attention feed-forward model with intra-sentence attention	90.5	No	A Decomposable Attention Model for Natural Langu...	2016-06-06	Code
45	200D decomposable attention model with intra-sentence attention	90.5	No	A Decomposable Attention Model for Natural Langu...	2016-06-06	Code
46	600D (300+300) Deep Gated Attn. BiLSTM encoders	90.5	No	Recurrent Neural Network-Based Sentence Encoder ...	2017-08-04	Code
47	600D Hierarchical BiLSTM with Max Pooling (HBMP, code)	89.9	No	Sentence Embeddings in NLI with Iterative Refine...	2018-08-27	Code
48	300D CAFE	89.8	No	Compare, Compress and Propagate: Enhancing Neura...	2017-12-30	-
49	300D Residual stacked encoders	89.8	No	Shortcut-Stacked Sentence Encoders for Multi-Dom...	2017-08-07	Code
50	Distance-based Self-Attention Network	89.6	No	Distance-based Self-Attention Network for Natura...	2017-12-06	-
51	200D decomposable attention feed-forward model	89.5	No	A Decomposable Attention Model for Natural Langu...	2016-06-06	Code
52	200D decomposable attention model	89.5	No	A Decomposable Attention Model for Natural Langu...	2016-06-06	Code
53	300D SPINN-PI encoders	89.2	No	A Fast Unified Model for Parsing and Sentence Un...	2016-03-19	Code
54	SLRC	89.1	No	Explicit Contextual Semantics for Text Comprehen...	2018-09-08	-
55	2400D Multiple-Dynamic Self-Attention Model	89	No	Dynamic Self-Attention : Computing Attention ove...	2018-08-22	Code
56	Biattentive Classification Network + CoVe + Char	88.5	No	Learned in Translation: Contextualized Word Vect...	2017-08-01	Code
57	300D Full tree matching NTI-SLSTM-LSTM w/ global attention	88.5	No	Neural Tree Indexers for Text Understanding	2016-07-15	Code
58	450D LSTMN with deep attention fusion	88.5	No	Long Short-Term Memory-Networks for Machine Read...	2016-01-25	Code
59	600D Dynamic Self-Attention Model	87.3	No	Dynamic Self-Attention : Computing Attention ove...	2018-08-22	Code
60	300D CAFE (no cross-sentence attention)	87.3	No	Compare, Compress and Propagate: Enhancing Neura...	2017-12-30	-
61	300D LSTMN with deep attention fusion	87.3	No	Long Short-Term Memory-Networks for Machine Read...	2016-01-25	Code
62	300D MMA-NSE encoders with attention	86.9	No	Neural Semantic Encoders	2016-07-14	Code
63	50D stacked TC-LSTMs	86.7	No	Modelling Interaction of Sentence Pair with coup...	2016-05-18	-
64	600D (300+300) BiLSTM encoders	86.4	No	Learning Natural Language Inference using Bidire...	2016-05-30	Code
65	300D NSE encoders	86.2	No	Neural Semantic Encoders	2016-07-14	Code
66	600D (300+300) BiLSTM encoders with intra-attention and symbolic preproc.	85.9	No	Learning Natural Language Inference using Bidire...	2016-05-30	Code
67	4096D BiLSTM with max-pooling	85.6	No	Supervised Learning of Universal Sentence Repres...	2017-05-05	Code
68	100D LSTMs w/ word-by-word attention	85.3	No	Reasoning about Entailment with Neural Attention	2015-09-22	Code
69	100D DF-LSTM	85.2	No	-	-	-
70	100D LSTM encoders	84.8	No	A large annotated corpus for learning natural la...	2015-08-21	Code
71	600D (300+300) BiLSTM encoders with intra-attention	84.5	No	Learning Natural Language Inference using Bidire...	2016-05-30	Code
72	300D LSTM encoders	83.9	No	A Fast Unified Model for Parsing and Sentence Un...	2016-03-19	Code
73	300D Tree-based CNN encoders	83.3	No	Natural Language Inference by Tree-Based Convolu...	2015-12-28	-
74	300D NTI-SLSTM-LSTM encoders	82.5	No	Neural Tree Indexers for Text Understanding	2016-07-15	Code
75	Unlexicalized features	49.4	No	A large annotated corpus for learning natural la...	2015-08-21	Code

#1+ Unigram and bigram featuresSOTA
99.7
% Train Accuracy· 2015-08-21
A large annotated corpus for learning natural language inference Code
#2Ntumpha
99.1
% Train Accuracy· 2019-01-31
Multi-Task Deep Neural Networks for Natural Language Understanding Code
#31024D GRU encoders w/ unsupervised 'skip-thoughts' pre-training
98.8
% Train Accuracy· 2015-11-19
Order-Embeddings of Images and Language Code
#4MT-DNN
97.2
% Train Accuracy· 2019-01-31
Multi-Task Deep Neural Networks for Natural Language Understanding Code
#5Fine-Tuned LM-Pretrained Transformer
96.6
% Train Accuracy
No paperCode
#6300D DMAN Ensemble
96.1
% Train Accuracy· 2019-07-23
Discourse Marker Augmented Network with Reinforcement Learning for Natural Language Inference Code
#7300D DMAN Ensemble
96.1
% Train Accuracy· 2019-07-23
Discourse Marker Augmented Network with Reinforcement Learning for Natural Language Inference Code
#8SJRC (BERT-Large +SRL)
95.7
% Train Accuracy· 2018-09-08
Explicit Contextual Semantics for Text Comprehension
#9150D Multiway Attention Network Ensemble
95.5
% Train Accuracy
No paperCode
#10300D DMAN
95.4
% Train Accuracy· 2019-07-23
Discourse Marker Augmented Network with Reinforcement Learning for Natural Language Inference Code
#11300D DMAN
95.4
% Train Accuracy· 2019-07-23
Discourse Marker Augmented Network with Reinforcement Learning for Natural Language Inference Code
#12Densely-Connected Recurrent and Co-Attentive Network Ensemble
95
% Train Accuracy· 2018-05-29
Semantic Sentence Matching with Densely-connected Recurrent and Co-attentive Information
#13600D BiLSTM with generalized pooling
94.9
% Train Accuracy· 2018-06-26
Enhancing Sentence Embedding with Generalized Pooling Code
#14450D DR-BiLSTM Ensemble
94.8
% Train Accuracy· 2018-02-15
DR-BiLSTM: Dependent Reading Bidirectional LSTM for Natural Language Inference
#15150D Multiway Attention Network
94.5
% Train Accuracy
No paperCode
#16SemBERT
94.4
% Train Accuracy· 2019-09-05
Semantics-aware BERT for Language Understanding Code
#17KIM
94.1
% Train Accuracy· 2017-11-12
Neural Natural Language Inference Models Enhanced with External Knowledge Code
#18450D DR-BiLSTM
94.1
% Train Accuracy· 2018-02-15
DR-BiLSTM: Dependent Reading Bidirectional LSTM for Natural Language Inference
#19RE2
94
% Train Accuracy· 2019-08-01
Simple and Effective Text Matching with Richer Alignment Features Code
#20KIM Ensemble
93.6
% Train Accuracy· 2017-11-12
Neural Natural Language Inference Models Enhanced with External Knowledge Code
#21600D ESIM + 300D Syntactic TreeLSTM
93.5
% Train Accuracy· 2016-09-20
Enhanced LSTM for Natural Language Inference Code
#22Stochastic Answer Network
93.3
% Train Accuracy· 2018-04-21
Stochastic Answer Networks for Natural Language Inference Code
#23BiMPM Ensemble
93.2
% Train Accuracy· 2017-02-13
Bilateral Multi-Perspective Matching for Natural Language Sentences Code
#24MFAE
93.18
% Train Accuracy
No paperCode
#25Densely-Connected Recurrent and Co-Attentive Network
93.1
% Train Accuracy· 2018-05-29
Semantic Sentence Matching with Densely-connected Recurrent and Co-attentive Information
#26600D Gumbel TreeLSTM encoders
93.1
% Train Accuracy· 2017-07-10
Learning to Compose Task-Specific Tree Structures Code
#27CA-MTL
92.6
% Train Accuracy· 2020-09-19
Conditionally Adaptive Multi-Task Learning: Improving Transfer Learning in NLP Using Fewer Parameters & Less Data Code
#28DEIM
92.6
% Train Accuracy· 2022-03-20
DEIM: An effective deep encoding and interaction model for sentence matching
#29300D Reinforced Self-Attention Network
92.6
% Train Accuracy· 2018-01-31
Reinforced Self-Attention Network: a Hybrid of Hard and Soft Attention for Sequence Modeling Code
#30300D CAFE Ensemble
92.5
% Train Accuracy· 2017-12-30
Compare, Compress and Propagate: Enhancing Neural Architectures with Alignment Factorization for Natural Language Inference
#31448D Densely Interactive Inference Network (DIIN, code) Ensemble
92.3
% Train Accuracy· 2017-09-13
Natural Language Inference over Interaction Space Code
#32ESIM + ELMo Ensemble
92.1
% Train Accuracy· 2018-02-15
Deep contextualized word representations Code
#33300D mLSTM word-by-word attention model
92
% Train Accuracy· 2015-12-30
Learning Natural Language Inference with LSTM Code
#34ESIM + ELMo
91.6
% Train Accuracy· 2018-02-15
Deep contextualized word representations Code
#35512D Dynamic Meta-Embeddings
91.6
% Train Accuracy· 2018-04-21
Dynamic Meta-Embeddings for Improved Sentence Representations Code
#36Densely-Connected Recurrent and Co-Attentive Network (encoder)
91.4
% Train Accuracy· 2018-05-29
Semantic Sentence Matching with Densely-connected Recurrent and Co-attentive Information
#37448D Densely Interactive Inference Network (DIIN, code)
91.2
% Train Accuracy· 2017-09-13
Natural Language Inference over Interaction Space Code
#38300D Gumbel TreeLSTM encoders
91.2
% Train Accuracy· 2017-07-10
Learning to Compose Task-Specific Tree Structures Code
#39300D Directional self-attention network encoders
91.1
% Train Accuracy· 2017-09-14
DiSAN: Directional Self-Attention Network for RNN/CNN-Free Language Understanding Code
#40600D Residual stacked encoders
91
% Train Accuracy· 2017-08-07
Shortcut-Stacked Sentence Encoders for Multi-Domain Inference Code
#41BiMPM
90.9
% Train Accuracy· 2017-02-13
Bilateral Multi-Perspective Matching for Natural Language Sentences Code
#42300D re-read LSTM
90.7
% Train Accuracy
No paper
#43300D re-read LSTM
90.7
% Train Accuracy
No paper
#44200D decomposable attention feed-forward model with intra-sentence attention
90.5
% Train Accuracy· 2016-06-06
A Decomposable Attention Model for Natural Language Inference Code
#45200D decomposable attention model with intra-sentence attention
90.5
% Train Accuracy· 2016-06-06
A Decomposable Attention Model for Natural Language Inference Code
#46600D (300+300) Deep Gated Attn. BiLSTM encoders
90.5
% Train Accuracy· 2017-08-04
Recurrent Neural Network-Based Sentence Encoder with Gated Attention for Natural Language Inference Code
#47600D Hierarchical BiLSTM with Max Pooling (HBMP, code)
89.9
% Train Accuracy· 2018-08-27
Sentence Embeddings in NLI with Iterative Refinement Encoders Code
#48300D CAFE
89.8
% Train Accuracy· 2017-12-30
Compare, Compress and Propagate: Enhancing Neural Architectures with Alignment Factorization for Natural Language Inference
#49300D Residual stacked encoders
89.8
% Train Accuracy· 2017-08-07
Shortcut-Stacked Sentence Encoders for Multi-Domain Inference Code
#50Distance-based Self-Attention Network
89.6
% Train Accuracy· 2017-12-06
Distance-based Self-Attention Network for Natural Language Inference
#51200D decomposable attention feed-forward model
89.5
% Train Accuracy· 2016-06-06
A Decomposable Attention Model for Natural Language Inference Code
#52200D decomposable attention model
89.5
% Train Accuracy· 2016-06-06
A Decomposable Attention Model for Natural Language Inference Code
#53300D SPINN-PI encoders
89.2
% Train Accuracy· 2016-03-19
A Fast Unified Model for Parsing and Sentence Understanding Code
#54SLRC
89.1
% Train Accuracy· 2018-09-08
Explicit Contextual Semantics for Text Comprehension
#552400D Multiple-Dynamic Self-Attention Model
89
% Train Accuracy· 2018-08-22
Dynamic Self-Attention : Computing Attention over Words Dynamically for Sentence Embedding Code
#56Biattentive Classification Network + CoVe + Char
88.5
% Train Accuracy· 2017-08-01
Learned in Translation: Contextualized Word Vectors Code
#57300D Full tree matching NTI-SLSTM-LSTM w/ global attention
88.5
% Train Accuracy· 2016-07-15
Neural Tree Indexers for Text Understanding Code
#58450D LSTMN with deep attention fusion
88.5
% Train Accuracy· 2016-01-25
Long Short-Term Memory-Networks for Machine Reading Code
#59600D Dynamic Self-Attention Model
87.3
% Train Accuracy· 2018-08-22
Dynamic Self-Attention : Computing Attention over Words Dynamically for Sentence Embedding Code
#60300D CAFE (no cross-sentence attention)
87.3
% Train Accuracy· 2017-12-30
Compare, Compress and Propagate: Enhancing Neural Architectures with Alignment Factorization for Natural Language Inference
#61300D LSTMN with deep attention fusion
87.3
% Train Accuracy· 2016-01-25
Long Short-Term Memory-Networks for Machine Reading Code
#62300D MMA-NSE encoders with attention
86.9
% Train Accuracy· 2016-07-14
Neural Semantic Encoders Code
#6350D stacked TC-LSTMs
86.7
% Train Accuracy· 2016-05-18
Modelling Interaction of Sentence Pair with coupled-LSTMs
#64600D (300+300) BiLSTM encoders
86.4
% Train Accuracy· 2016-05-30
Learning Natural Language Inference using Bidirectional LSTM model and Inner-Attention Code
#65300D NSE encoders
86.2
% Train Accuracy· 2016-07-14
Neural Semantic Encoders Code
#66600D (300+300) BiLSTM encoders with intra-attention and symbolic preproc.
85.9
% Train Accuracy· 2016-05-30
Learning Natural Language Inference using Bidirectional LSTM model and Inner-Attention Code
#674096D BiLSTM with max-pooling
85.6
% Train Accuracy· 2017-05-05
Supervised Learning of Universal Sentence Representations from Natural Language Inference Data Code
#68100D LSTMs w/ word-by-word attention
85.3
% Train Accuracy· 2015-09-22
Reasoning about Entailment with Neural Attention Code
#69100D DF-LSTM
85.2
% Train Accuracy
No paper
#70100D LSTM encoders
84.8
% Train Accuracy· 2015-08-21
A large annotated corpus for learning natural language inference Code
#71600D (300+300) BiLSTM encoders with intra-attention
84.5
% Train Accuracy· 2016-05-30
Learning Natural Language Inference using Bidirectional LSTM model and Inner-Attention Code
#72300D LSTM encoders
83.9
% Train Accuracy· 2016-03-19
A Fast Unified Model for Parsing and Sentence Understanding Code
#73300D Tree-based CNN encoders
83.3
% Train Accuracy· 2015-12-28
Natural Language Inference by Tree-Based Convolution and Heuristic Matching
#74300D NTI-SLSTM-LSTM encoders
82.5
% Train Accuracy· 2016-07-15
Neural Tree Indexers for Text Understanding Code
#75Unlexicalized features
49.4
% Train Accuracy· 2015-08-21
A large annotated corpus for learning natural language inference Code