Browse › Natural Language Processing › Natural Language Inference › SNLI

Natural Language Inference archive 2025-07-28

SNLI Benchmark (Natural Language Inference)

98 rows 73 with code listed 6 metrics Dataset page

Syntology Since the archive (Syntology): none placed; 4 newer papers judged to report on this board so far.

Natural language inference (NLI) is the task of determining whether a "hypothesis" is true (entailment), false (contradiction), or undetermined (neutral) given a "premise".

Example:

Premise Label Hypothesis
A man inspects the uniform of a figure in some East Asian country. contradiction The man is sleeping.
An older and younger man smiling. neutral Two men are smiling and laughing at the cats playing on the floor.
A soccer game with multiple males playing. entailment Some men are playing a sport.

Approaches used for NLI include earlier symbolic and statistical approaches to more recent deep learning approaches. Benchmark datasets used for NLI include SNLI, MultiNLI, SciTail, among others. You can get hands-on practice on the SNLI task by following this d2l.ai chapter.

Further readings:

The archive carries no text for this table; the description above is the archive's text for the task Natural Language Inference. archive 2025-07-28

Over time archive 2025-07-28

Direction inferred from the metric name, not from the archive: % Test Accuracy (higher is better), % Train Accuracy (higher is better), Dev Accuracy (higher is better), % Dev Accuracy (higher is better), Accuracy (higher is better). Not inferred (points only, no best-so-far line): Parameters. Points are placed at the row's paper date, not an evaluation date; 94 of 98 rows carry one.

The chart needs JavaScript; the table below carries every value.

Results archive 2025-07-28

Archive rows end at the archive snapshot, 2025-07-28: no result published after that date is in this table. Rank is the archive's row order at that snapshot; not re-ranked here. Metric values are the archive's strings. Column headers sort the table in your browser; each row keeps its archive rank.

Paper Code Ran Syntology Report
1 UnitedSynT5 (3B) 94.7 ✓ Paper – 2024 no code linked report
2 UnitedSynT5 (335M) 93.5 ✓ Paper – 2024 no code linked report
3 Neural Tree Indexers for Text Understanding 93.1355 – Paper Code 2021 1 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 1 violated, 0 with no contract checked; 0 where Syntology's instrument failed) · 2 unverified report
4 EFL (Entailment as Few-shot Learner) + RoBERTa-large 93.1?355m – Paper Code 2021 1 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 1 violated, 0 with no contract checked; 0 where Syntology's instrument failed) · 2 unverified report
5 RoBERTa-large+Self-Explaining 92.3340 – Paper Code 2020 linked, not harvested report
6 RoBERTa-large + self-explaining layer 92.3?355m+ – Paper Code 2020 linked, not harvested report
7 CA-MTL 92.192.6340m – Paper Code 2020 linked, not harvested report
8 SemBERT 91.994.4339m – Paper Code 2019 official (archive's flag): 7 ran · 7 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 2 honoured, 0 violated, 3 with no contract checked; 2 where Syntology's instrument failed) · 5 unverified report
9 MT-DNN-SMARTLARGEv0 91.792.6 – Paper Code 2019 official: no sample here; runs from other or unrecorded repositories · 8 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 2 honoured, 1 violated, 2 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified report
10 MT-DNN 91.697.2330m – Paper Code 2019 community repositories only · 9 ran (of which 0 constructed an object rather than computing a result; 4 with no instrument failure: 1 honoured, 1 violated, 2 with no contract checked; 5 where Syntology's instrument failed) · 4 unverified report
11 SJRC (BERT-Large +SRL) 91.395.7308m – Paper – 2018 no code linked report
12 Ntumpha 90.599.1220 – Paper Code 2019 community repositories only · 9 ran (of which 0 constructed an object rather than computing a result; 4 with no instrument failure: 1 honoured, 1 violated, 2 with no contract checked; 5 where Syntology's instrument failed) · 4 unverified report
13 Densely-Connected Recurrent and Co-Attentive Network Ensemble 90.195.053.3m – Paper – 2018 no code linked report
14 MFAE 90.0793.18 – Paper Code 2020 linked, not harvested report
15 Fine-Tuned LM-Pretrained Transformer 89.996.685m – Paper Code 2018 linked, not harvested report
16 300D DMAN Ensemble 89.696.179m – Paper Code 2019 linked, not harvested report
17 300D DMAN Ensemble 89.696.179m – – – not matched report
18 150D Multiway Attention Network Ensemble 89.495.558m – Paper Code 2018 linked, not harvested report
19 450D DR-BiLSTM Ensemble 89.394.845m – Paper – 2018 no code linked report
20 300D CAFE Ensemble 89.392.517.5m – Paper – 2017 no code linked report
21 ESIM + ELMo Ensemble 89.392.140m – Paper Code 2018 30 ran (of which 12 constructed an object rather than computing a result; 24 with no instrument failure: 2 honoured, 0 violated, 22 with no contract checked; 6 where Syntology's instrument failed) · 28 unverified report
22 KIM Ensemble 89.193.643m – Paper Code 2017 linked, not harvested report
23 SLRC 89.189.16.1m – Paper – 2018 no code linked report
24 RE2 88.994.02.8m – Paper Code 2019 community repositories only · 1 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified report
25 Densely-Connected Recurrent and Co-Attentive Network 88.993.16.7m – Paper – 2018 no code linked report
26 DEIM 88.992.622m – Paper – 2022 no code linked report
27 448D Densely Interactive Inference Network (DIIN, code) Ensemble 88.992.317m – Paper Code 2017 community repositories only · 1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified report
28 300D DMAN 88.895.49.2m – Paper Code 2019 linked, not harvested report
29 300D DMAN 88.895.49.2m – – – not matched report
30 BiMPM Ensemble 88.893.26.4m – Paper Code 2017 1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 3 unverified report
31 ESIM + ELMo 88.791.68.0m – Paper Code 2018 30 ran (of which 12 constructed an object rather than computing a result; 24 with no instrument failure: 2 honoured, 0 violated, 22 with no contract checked; 6 where Syntology's instrument failed) · 28 unverified report
32 KIM 88.694.14.3m – Paper Code 2017 linked, not harvested report
33 600D ESIM + 300D Syntactic TreeLSTM 88.693.57.7m – Paper Code 2016 linked, not harvested report
34 450D DR-BiLSTM 88.594.17.5m – Paper – 2018 no code linked report
35 Stochastic Answer Network 88.593.33.5m – Paper Code 2018 linked, not harvested report
36 300D CAFE 88.589.84.7m – Paper – 2017 no code linked report
37 150D Multiway Attention Network 88.394.514m – Paper Code 2018 linked, not harvested report
38 Biattentive Classification Network + CoVe + Char 88.188.522m – Paper Code 2017 community repositories only · 3 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 2 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified report
39 aESIM 88.1 – Paper – 2018 no code linked report
40 448D Densely Interactive Inference Network (DIIN, code) 88.091.24.4m – Paper Code 2017 community repositories only · 1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified report
41 Enhanced Sequential Inference Model (Chen et al., [2017a]) 88.0 – Paper Code 2016 linked, not harvested report
42 BiMPM 87.590.91.6m – Paper Code 2017 1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 3 unverified report
43 300D re-read LSTM 87.590.72.0m – – – not matched report
44 300D re-read LSTM 87.590.72.0m – Paper – 2016 no code linked report
45 2400D Multiple-Dynamic Self-Attention Model 87.489.07.0m – Paper Code 2018 linked, not harvested report
46 300D Full tree matching NTI-SLSTM-LSTM w/ global attention 87.388.53.2m – Paper Code 2016 linked, not harvested report
47 300D 2-layer Bi-CAS-LSTM 87 – Paper – 2018 no code linked report
48 200D decomposable attention feed-forward model with intra-sentence attention 86.890.5580k – Paper Code 2016 4 ran (of which 0 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 0 where Syntology's instrument failed) · 2 unverified report
49 200D decomposable attention model with intra-sentence attention 86.890.5580k – Paper Code 2016 4 ran (of which 0 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 0 where Syntology's instrument failed) · 2 unverified report
50 600D Dynamic Self-Attention Model 86.887.32.1m – Paper Code 2018 linked, not harvested report
51 CBS-1 + ESIM 86.73 – Paper – 2018 no code linked report
52 512D Dynamic Meta-Embeddings 86.791.69m – Paper Code 2018 linked, not harvested report
53 600D BiLSTM with generalized pooling 86.694.965m – Paper Code 2018 linked, not harvested report
54 600D Hierarchical BiLSTM with Max Pooling (HBMP, code) 86.689.922m – Paper Code 2018 linked, not harvested report
55 Densely-Connected Recurrent and Co-Attentive Network (encoder) 86.591.45.6m – Paper – 2018 no code linked report
56 300D Reinforced Self-Attention Network 86.392.63.1m – Paper Code 2018 linked, not harvested report
57 Distance-based Self-Attention Network 86.389.64.7m – Paper – 2017 no code linked report
58 200D decomposable attention feed-forward model 86.389.5380k – Paper Code 2016 4 ran (of which 0 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 0 where Syntology's instrument failed) · 2 unverified report
59 200D decomposable attention model 86.389.5380k – Paper Code 2016 4 ran (of which 0 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 0 where Syntology's instrument failed) · 2 unverified report
60 450D LSTMN with deep attention fusion 86.388.53.4m – Paper Code 2016 2 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified report
61 300D mLSTM word-by-word attention model 86.192.01.9m – Paper Code 2015 community repositories only · 1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified report
62 600D Gumbel TreeLSTM encoders 86.093.110m – Paper Code 2017 linked, not harvested report
63 600D Residual stacked encoders 86.091.029m – Paper Code 2017 linked, not harvested report
64 Star-Transformer (no cross sentence attention) 86.0 – Paper Code 2019 official: harvested for another paper · 0 ran · 1 unverified report
65 300D CAFE (no cross-sentence attention) 85.987.33.7m – Paper – 2017 no code linked report
66 1200D REGMAPR (Base+Reg) 85.9–– – – – not matched report
67 300D Residual stacked encoders 85.789.89.7m – Paper Code 2017 linked, not harvested report
68 300D LSTMN with deep attention fusion 85.787.31.7m – Paper Code 2016 2 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified report
69 300D Gumbel TreeLSTM encoders 85.691.22.9m – Paper Code 2017 linked, not harvested report
70 300D Directional self-attention network encoders 85.691.12.4m – Paper Code 2017 linked, not harvested report
71 600D (300+300) Deep Gated Attn. BiLSTM encoders 85.590.512m – Paper Code 2017 linked, not harvested report
72 300D MMA-NSE encoders with attention 85.486.93.2m – Paper Code 2016 linked, not harvested report
73 50D stacked TC-LSTMs 85.186.7190k – Paper – 2016 no code linked report
74 600D (300+300) BiLSTM encoders with intra-attention and symbolic preproc. 85.085.92.8m – Paper Code 2016 linked, not harvested report
75 Stacked Bi-LSTMs (shortcut connections, max-pooling) 84.8 – Paper Code 2018 linked, not harvested report
76 300D NSE encoders 84.686.23.0m – Paper Code 2016 linked, not harvested report
77 100D DF-LSTM 84.685.2320k – Paper – 2016 no code linked report
78 4096D BiLSTM with max-pooling 84.585.640m – Paper Code 2017 official (archive's flag): 3 ran · 6 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 2 honoured, 0 violated, 0 with no contract checked; 4 where Syntology's instrument failed) · 1 unverified report
79 Bi-LSTM sentence encoder (max-pooling) 84.5 – Paper Code 2018 linked, not harvested report
80 Stacked Bi-LSTMs (shortcut connections, max-pooling, attention) 84.4 – Paper Code 2018 linked, not harvested report
81 600D (300+300) BiLSTM encoders with intra-attention 84.284.52.8m – Paper Code 2016 linked, not harvested report
82 SWEM-max 83.8 – Paper Code 2018 linked, not harvested report
83 100D LSTMs w/ word-by-word attention 83.585.3250k – Paper Code 2015 1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified report
84 300D NTI-SLSTM-LSTM encoders 83.482.54.0m – Paper Code 2016 linked, not harvested report
85 600D (300+300) BiLSTM encoders 83.386.42.0m – Paper Code 2016 linked, not harvested report
86 300D SPINN-PI encoders 83.289.23.7m – Paper Code 2016 official (archive's flag): 3 ran · 6 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 0 honoured, 0 violated, 6 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified report
87 300D Tree-based CNN encoders 82.183.33.5m – Paper – 2015 no code linked report
88 1024D GRU encoders w/ unsupervised 'skip-thoughts' pre-training 81.498.815m – Paper Code 2015 linked, not harvested report
89 DELTA (LSTM) 80.7 – Paper Code 2019 linked, not harvested report
90 300D LSTM encoders 80.683.93.0m – Paper Code 2016 official (archive's flag): 3 ran · 6 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 0 honoured, 0 violated, 6 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified report
91 + Unigram and bigram features 78.299.7 – Paper Code 2015 3 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 1 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified report
92 100D LSTM encoders 77.684.8220k – Paper Code 2015 3 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 1 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified report
93 Unlexicalized features 50.449.4 – Paper Code 2015 3 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 1 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified report
94 MT-DNN-SMART_100%ofTrainingData 91.6 – Paper Code 2019 official: no sample here; runs from other or unrecorded repositories · 8 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 2 honoured, 1 violated, 2 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified report
95 MT-DNN-SMART_10%ofTrainingData 88.7 – Paper Code 2019 official: no sample here; runs from other or unrecorded repositories · 8 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 2 honoured, 1 violated, 2 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified report
96 MT-DNN-SMART_1%ofTrainingData 86 – Paper Code 2019 official: no sample here; runs from other or unrecorded repositories · 8 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 2 honoured, 1 violated, 2 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified report
97 MT-DNN-SMART_0.1%ofTrainingData 82.7 – Paper Code 2019 official: no sample here; runs from other or unrecorded repositories · 8 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 2 honoured, 1 violated, 2 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified report
98 SplitEE-S 79.0 – Paper Code 2023 linked, not harvested report

All 98 rows shown. 94 link to a paper page on this site; 2 are marked as using additional training data in the archive. No GitHub stars are tracked; "Code" is the first repository the archive lists for the row. The archive carries no row tags, review links or community-submitted rows for this table; none are shown. archive 2025-07-28

Syntology Ran reads “N ran (of which C constructed an object rather than computing a result; K with no instrument failure: H honoured, V violated, P with no contract checked; I where Syntology's instrument failed) · U unverified”: of the code samples Syntology harvested from repositories linked to that row's paper (joined by arXiv id), N executed on a synthesized input and U are unverified (harvested, no recorded run). Of the N, C constructed an object rather than computing a result (a part of N, never taken away from it), K ran with no instrument failure (honoured or violated the contract Syntology drafted, or ran with no contract checked) and I are failures of Syntology's instrument, not of the code (a wrong contract draft, or a contract check that could not run). It counts code from repositories linked to that row's paper, not this result: the row's number was not reproduced and nothing here is a correctness claim. The other cell texts mean no graph line for the row: "linked, not harvested" (the archive links code, Syntology has not harvested it), "no code linked" (no code link in the archive), "not matched" (the row's paper URL matched no paper on this site). 33 rows have a graph line, from 17 distinct papers. Rows whose paper has a sample that ran: 32 (26 with a run with no instrument failure, 6 where every run was a failure of Syntology's instrument). Papers: 16 with a sample that ran: 12 with a run with no instrument failure, 4 where every run was a failure of Syntology's instrument. Whose code ran. When the archive marks a repository official for the row's paper, the cell starts with that repository's state (“official” is the archive's flag, not a verdict on who wrote the code): “official (archive's flag): N ran” counts the samples from it that ran, instrument failures included; “official: harvested, nothing ran”; when it has no sample for this paper and other samples ran, “community repositories only” if every one came from a repository the archive lists for the paper but neither marks official nor records as mentioned in it, or one Syntology harvested outside the archive's links, and “official: no sample here; runs from other or unrecorded repositories” if any came from a repository named in the paper or found in its text by Syntology, or has no repository recorded; “official: harvested for another paper” (no sample from it is linked to this paper) or “official: not harvested”. No state is printed when the archive marks none official. The cell's tooltip names the repositories the samples that ran came from. Of the 33 rows with a graph line, 17 have an archive-official repository: on 4 a sample from it ran (4 with a run with no instrument failure), on 7 the samples that ran came from community repositories only, and on 5 from other or unrecorded repositories. The filters above hide rows and never renumber them; the first two count, by default, only a run with no instrument failure, and offer any run, instrument failures included. Counting each paper once, of 130 samples: 84 ran (of which 12 constructed an object rather than computing a result; 56 with no instrument failure: 13 honoured, 4 violated, 39 with no contract checked; 28 where Syntology's instrument failed) · 46 unverified. Separately, 51 of those 130 are pointer-only (licence): the site points at that code rather than redistributing it, a licence property recorded for ran and unverified samples alike; each cell's tooltip carries the row's own pointer-only count. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. Per-sample status is on the paper page; each Ran cell links to that paper's sample list. For agents: get_harvested_code_for_paper(arxiv_id) lists each paper's samples; how to connect.

Since the archive Syntology

Of the 8,888 newer papers whose tables Syntology has read (8,886 checked in full and 2 with a leaderboard still awaiting a verdict), 11 had this board among the leaderboards the extractor considered for their tables (it keeps at most 10 boards per paper and 4 per dataset; the lists of 1,586 of the 8,888 were already full without this board, so if it was crowded out of them it is not counted), and 4 proposed a result; of those 4, 4 gave no cell that passed the extractor's own checks, 0 were rejected by the independent check, 0 await a verdict and 0 were placed. Syntology has checked 8,886 of the 18,637 papers on this site that are newer than the archive (for 3,717 of them no archive leaderboard matched the paper's tables, so there was nothing further to check); 733 were read and have nothing to place (arXiv has no HTML version of the paper, or that version has no tables), 41 could not be read (the extractor's reply could not be parsed), and results from the other 8,977 appear after they are checked.

Newer papers whose tables Claude Sonnet 4.5 judged to report on this benchmark (a model's judgement, not a result; newest first by arXiv id)

Paper arXiv date Ran Syntology Outcome
Quantized Evolution Strategies: High-precision Fine-tuning of Quantized LLMs at... arXiv:2602.03120 3 Feb 2026 9 harvested: 9 ran (of which 0 constructed an object rather than computing a result; 7 with no instrument failure: 0 honoured, 0 violated, 7 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified refused by a rule: the table's column header does not name the leaderboard's metric
Stable Coresets via Posterior Sampling: Aligning Induced and Full Loss Landscapes arXiv:2511.17399 21 Nov 2025 31 harvested: 17 ran (of which 0 constructed an object rather than computing a result; 8 with no instrument failure: 0 honoured, 2 violated, 6 with no contract checked; 9 where Syntology's instrument failed) · 14 unverified refused by a rule: the value is not on the scale of the leaderboard column's archive values
Accurate and Efficient Low-Rank Model Merging in Core Space arXiv:2509.17786 22 Sep 2025 5 harvested: 5 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 4 where Syntology's instrument failed) · 0 unverified refused by a rule: the table's column header does not name the leaderboard's metric
CoBA: Counterbias Text Augmentation for Mitigating Various Spurious Correlations via... arXiv:2508.21083 26 Aug 2025 repository linked, no samples harvested refused by a rule: the table's column header does not name the leaderboard's metric

Syntology 4 newer papers judged to report on this board: the papers counted above as having proposed a result (0 placed); every one is listed here and in the JSON twin. The order is the arXiv identifier's, newest first, not a rank. The judgement is the model's, read from the paper's own tables; the outcome is what Syntology's checks did with it, and no value from a paper that was not placed is printed. “refused by a rule” names the extractor's rule; “no cell passed the extractor's checks” means the extractor's record holds no cell for this board, passed or refused. The 4 counted above as giving no cell that passed the extractor's own checks are the 4 marked “refused by a rule”. Drawn from 8,886 of 18,637 newer papers checked so far (733 more had nothing to place; 41 could not be read).

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections