Browse › Natural Language Processing › Natural Language Inference › CommitmentBank
CommitmentBank Benchmark (Natural Language Inference)
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), F1 (higher is better). Points are placed at the row's paper date; 20 of 20 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 | PaLM 540B (finetuned) | 100 | 100 | – | Paper | Code | 2022 | 30 of 37 ran · 7 unverified | report |
| 2 | Vega v2 6B (KD-based prompt transfer) | 99.2 | 98.6 | – | Paper | – | 2022 | no code linked | report |
| 3 | ST-MoE-L 4.1B (fine-tuned) | 98.2 | – | Paper | Code | 2022 | 5 of 5 ran · 0 unverified | report | |
| 4 | ST-MoE-32B 269B (fine-tuned) | 98 | – | Paper | Code | 2022 | 5 of 5 ran · 0 unverified | report | |
| 5 | Turing NLR v5 XXL 5.4B (fine-tuned) | 97.6 | 95.9 | – | Paper | – | 2022 | no code linked | report |
| 6 | DeBERTa-1.5B | 97.2 | 94.9 | – | Paper | Code | 2020 | 4 of 13 ran · 9 unverified | report |
| 7 | T5-XXL 11B (fine-tuned) | 96.8 | 93.9 | – | Paper | Code | 2019 | 2 of 31 ran · 29 unverified | report |
| 8 | T5-Large 770M (fine-tuned) | 94.4 | 90.3 | – | Paper | Code | 2019 | 2 of 31 ran · 29 unverified | report |
| 9 | T5-Base 220M (fine-tuned) | 94 | 86.2 | – | Paper | Code | 2019 | 2 of 31 ran · 29 unverified | report |
| 10 | PaLM 2-L (one-shot) | 87.5 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 11 | PaLM 2-S (one-shot) | 82.1 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 12 | PaLM 2-M (one-shot) | 80.4 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 13 | GPT-3 175B (Few-Shot) | 75.6 | – | Paper | Code | 2020 | 15 of 65 ran · 50 unverified | report | |
| 14 | N-Grammer 343M | 67.9 | 59.7 | – | Paper | Code | 2022 | 0 of 6 ran · 6 unverified | report |
| 15 | AlexaTM 20B | 67.9 | – | Paper | Code | 2022 | 1 of 1 ran · 0 unverified | report | |
| 16 | Bloomberg GPT (one-shot) | 53.57 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 17 | GPT-NeoX (one-shot) | 48.21 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 18 | BLOOM 176B (one-shot) | 48.21 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 19 | OPT 66B (one-shot) | 44.64 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 20 | GPT-3 175B (few-shot, k=32) | 52 | – | Paper | Code | 2020 | 15 of 65 ran · 50 unverified | report |
All 20 rows shown. 20 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). 11 rows have a graph line, from 7 distinct papers; 10 rows (6 papers) have at least one sample that ran. Counting each paper once: Syntology ran 57 of 158 samples; 101 unverified. Separately, 12 of those 158 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