Browse › Natural Language Processing › Natural Language Inference › RTE

Natural Language Inference archive 2025-07-28

RTE Benchmark (Natural Language Inference)

90 rows 78 with code listed 1 metric Dataset page

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

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

Direction inferred from the metric name, not from the archive: Accuracy (higher is better). Points are placed at the row's paper date; 89 of 90 rows carry one.

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 Vega v2 6B (KD-based prompt transfer) 96% – Paper – 2022 no code linked report
2 PaLM 540B (fine-tuned) 95.7% – Paper Code 2022 30 of 37 ran · 7 unverified report
3 Turing NLR v5 XXL 5.4B (fine-tuned) 94.1% – Paper – 2022 no code linked report
4 ST-MoE-32B 269B (fine-tuned) 93.5% – Paper Code 2022 5 of 5 ran · 0 unverified report
5 DeBERTa-1.5B 93.2% – Paper Code 2020 4 of 13 ran · 9 unverified report
6 MUPPET Roberta Large 92.8% – Paper Code 2021 linked, not harvested report
7 DeBERTaV3large 92.7% – Paper Code 2021 0 of 7 ran · 7 unverified report
8 T5-XXL 11B 92.5% – Paper Code 2019 6 of 8 ran · 2 unverified report
9 T5-XXL 11B (fine-tuned) 92.5% – Paper Code 2019 2 of 31 ran · 29 unverified report
10 ST-MoE-L 4.1B (fine-tuned) 92.1% – Paper Code 2022 5 of 5 ran · 0 unverified report
11 UL2 20B (fine-tuned) 92.1% – Paper Code 2022 0 of 16 ran · 16 unverified report
12 SMARTRoBERTa 92.0% – Paper Code 2019 6 of 8 ran · 2 unverified report
13 FLAN 137B (prompt-tuned) 91.7% – Paper Code 2021 0 of 1 ran · 1 unverified report
14 T5-XL 3B 91.1% – Paper Code 2019 2 of 31 ran · 29 unverified report
15 RoBERTa-large 355M + Entailment as Few-shot Learner 90.5% – Paper Code 2021 1 of 3 ran · 2 unverified report
16 ALBERT 89.2% – Paper Code 2019 46 of 126 ran · 80 unverified report
17 Adv-RoBERTa ensemble 88.7% – Paper – 2019 no code linked report
18 RoBERTa 88.2% – Paper Code 2019 22 of 48 ran · 26 unverified report
19 RoBERTa (ensemble) 88.2% – Paper Code 2019 22 of 48 ran · 26 unverified report
20 T5-Large 738M 87.4% – Paper Code 2023 linked, not harvested report
21 T5-Large 770M 87.2% – Paper Code 2019 2 of 31 ran · 29 unverified report
22 RoBERTa-large 355M + EFL + UCA 87.2% – Paper Code 2021 1 of 3 ran · 2 unverified report
23 PSQ (Chen et al., 2020) 86.8 – Paper Code 2020 1 of 4 ran · 3 unverified report
24 XLNet (single model) 85.9% – Paper Code 2019 10 of 24 ran · 14 unverified report
25 RoBERTa-large 355M (MLP quantized vector-wise, fine-tuned) 85.4% – Paper Code 2022 2 of 5 ran · 3 unverified report
26 OPT-IML 175B 84.8% – Paper Code 2022 linked, not harvested report
27 Q8BERT (Zafrir et al., 2019) 84.8 – Paper Code 2019 3 of 11 ran · 8 unverified report
28 Q-BERT (Shen et al., 2020) 84.7 – Paper – 2019 no code linked report
29 FLAN 137B (8-shot) 84.5% – Paper Code 2021 0 of 1 ran · 1 unverified report
30 FLAN 137B (0-shot) 84.1% – Paper Code 2021 0 of 1 ran · 1 unverified report
31 OPT-IML 30B 83.8% – Paper Code 2022 linked, not harvested report
32 ELECTRA 83.6% – – – not matched report
33 PaLM 2-M (1-shot) 81.9% – Paper Code 2023 linked, not harvested report
34 T0-3B (CoT fine-tuned) 80.8% – Paper Code 2023 linked, not harvested report
35 ERNIE 2.0 Large 80.2% – Paper Code 2019 0 of 1 ran · 1 unverified report
36 T5-Base 220M 80.1% – Paper Code 2019 2 of 31 ran · 29 unverified report
37 MLM+ del-span 79.8% – Paper – 2020 no code linked report
38 PaLM 540B (5-shot) 79.6% – Paper Code 2022 30 of 37 ran · 7 unverified report
39 PaLM 2-L (1-shot) 79.3% – Paper Code 2023 linked, not harvested report
40 SpanBERT 79.0% – Paper Code 2019 3 of 15 ran · 12 unverified report
41 PaLM 2-S (1-shot) 78.7% – Paper Code 2023 linked, not harvested report
42 PaLM 540B (1-shot) 78.7% – Paper Code 2022 30 of 37 ran · 7 unverified report
43 Neo-6B (QA + WS) 75.1% – Paper Code 2022 2 of 2 ran · 0 unverified report
44 BigBird 75.0% – Paper Code 2020 10 of 15 ran · 5 unverified report
45 ERNIE 2.0 Base 74.8% – Paper Code 2019 0 of 1 ran · 1 unverified report
46 KiC-770M 74.00 – Paper – 2022 no code linked report
47 RealFormer 73.7% – Paper Code 2020 linked, not harvested report
48 SqueezeBERT 73.2% – Paper Code 2020 0 of 1 ran · 1 unverified report
49 PaLM 540B (0-shot) 72.9% – Paper Code 2022 30 of 37 ran · 7 unverified report
50 SMART-BERT 71.2% – Paper Code 2019 6 of 8 ran · 2 unverified report
51 SMART 71.2% – Paper Code 2019 6 of 8 ran · 2 unverified report
52 Flipped-3B 71.05 – Paper Code 2022 linked, not harvested report
53 BERT-large 340M 70.1% – Paper Code 2018 204 of 659 ran · 455 unverified report
54 T5-Small 69.9% – Paper Code 2019 2 of 31 ran · 29 unverified report
55 data2vec 69.9% – Paper Code 2022 0 of 6 ran · 6 unverified report
56 Bloomberg GPT 50B (1-shot) 69.3% – Paper Code 2023 linked, not harvested report
57 FNet-Large 69% – Paper Code 2021 2 of 2 ran · 0 unverified report
58 GPT-3 175B (few-shot, k=32) 69% – Paper Code 2020 15 of 65 ran · 50 unverified report
59 ERNIE 68.8% – Paper Code 2019 3 of 3 ran · 0 unverified report
60 AlexaTM 20B 68.6% – Paper Code 2022 1 of 1 ran · 0 unverified report
61 LaMini-GPT 1.5B 67.9% – Paper Code 2023 linked, not harvested report
62 SenseBERT-base 110M 67.5% – Paper – 2019 no code linked report
63 OPT-IML 1.3B 66.8% – Paper Code 2022 linked, not harvested report
64 TinyBERT-6 67M 66% – Paper Code 2019 0 of 4 ran · 4 unverified report
65 LaMini-F-T5 783M 65% – Paper Code 2023 linked, not harvested report
66 RoE-3B 64.01 – Paper Code 2023 3 of 3 ran · 0 unverified report
67 ELC-BERT-base 98M (zero init) 63 – Paper – 2023 no code linked report
68 DistilBERT 66M 62.9% – Paper Code 2019 19 of 27 ran · 8 unverified report
69 TinyBERT-4 14.5M 62.9% – Paper Code 2019 0 of 4 ran · 4 unverified report
70 Neo-6B (QA) 61.7% – Paper Code 2022 2 of 2 ran · 0 unverified report
71 UL2 20B (0-shot) 60.7% – Paper Code 2022 0 of 16 ran · 16 unverified report
72 OPT 175B 60.3% – Paper Code 2022 linked, not harvested report
73 N-Grammer 343M 59.2% – Paper Code 2022 0 of 6 ran · 6 unverified report
74 Hybrid H3 125M (0-shot, logit scoring) 59.2% – Paper Code 2022 7 of 15 ran · 8 unverified report
75 Neo-6B (few-shot) 58.8% – Paper Code 2022 2 of 2 ran · 0 unverified report
76 OPT 30B 58.1% – Paper Code 2022 linked, not harvested report
77 Hybrid H3 125M (3-shot, logit scoring) 58.1% – Paper Code 2022 7 of 15 ran · 8 unverified report
78 Hybrid H3 125M (3-shot, rank classification) 58.1% – Paper Code 2022 7 of 15 ran · 8 unverified report
79 24hBERT 57.7% – Paper Code 2021 linked, not harvested report
80 BLOOM 176B (1-shot) 57.4% – Paper Code 2023 linked, not harvested report
81 LaMini-T5 738M 57% – Paper Code 2023 linked, not harvested report
82 ELC-BERT-small 24M 55.4 – Paper – 2023 no code linked report
83 OPT 66B (1-shot) 54.9% – Paper Code 2023 linked, not harvested report
84 LTG-BERT-base 98M 54.7 – Paper – 2023 no code linked report
85 OPT 1.3B 54.2% – Paper Code 2022 linked, not harvested report
86 GPT-NeoX 20B (1-shot) 53.8% – Paper Code 2023 linked, not harvested report
87 LTG-BERT-small 24M 53.7 – Paper – 2023 no code linked report
88 H3 125M (0-shot, rank classification) 53.1% – Paper Code 2022 7 of 15 ran · 8 unverified report
89 GPT-2-XL 1.5B 52.3% – Paper Code 2023 linked, not harvested report
90 H3 125M (3-shot, rank classification) 52.3% – Paper Code 2022 7 of 15 ran · 8 unverified report

All 90 rows shown. 89 link to a paper page on this site; 0 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 of M ran · U unverified": of the M code samples Syntology harvested from repositories linked to that row's paper (joined by arXiv id), N executed on a synthesized input and the other U = M−N are unverified (harvested, no recorded 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). 55 rows have a graph line, from 31 distinct papers; 42 rows (23 papers) have at least one sample that ran. Counting each paper once: Syntology ran 401 of 1,164 samples; 763 unverified. Separately, 239 of those 1,164 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. Read from the graph 2026-09-24. Per-sample status is on the paper page.

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