The dataset viewer should be available soon. Please retry later.
OpenNLI eval sets
44 public test sets used for the OpenNLI leaderboard
Scoring script
evaluate.py (in this repo) reproduces the leaderboard numbers for sentence-transformers CrossEncoders
pip install "sentence-transformers>=5" datasets pandas
python evaluate.py --model cross-encoder/nli-deberta-v3-base --out results.json
python evaluate.py --model dleemiller/finecat-nli-ultra-m --trust-remote-code
Labels are mapped from the model's id2label, or use --label-order (e.g. ent,neu,con)
Scoring
- accuracy per dataset
- category score = mean over datasets
- overall = mean of the 4 category scores.
- Skill Index (the leaderboard headline) = Brier skill vs a uniform guess, per dataset (floored at −100), averaged the same way, at each model's fitted temperature (see Calibration).
For license information, refer to the original dataset's license.
DocNLI (1,000) and VitaminC (10,000) are fixed random samples of their test splits.
Calibration
The calibration config (split validation, 11,018 rows) is used only to fit each model's temperature: one number that
rescales the model's confidence before the Skill Index is computed. It's drawn from the validation / dev splits of 32
benchmarks (never a test split), up to 400 rows each, labelled like the test sets, with no text overlapping any test set.
Models may calibrate on it however they like; it never scores anything.
| Config | Dataset | Category | Rows | Labels |
|---|---|---|---|---|
mnli |
MNLI-m | NLI | 9,815 | 3way |
mnli_mismatched |
MNLI-mm | NLI | 9,830 | 3way |
snli |
SNLI | NLI | 9,824 | 3way |
anli_r1 |
ANLI R1 | NLI | 1,000 | 3way |
anli_r2 |
ANLI R2 | NLI | 1,000 | 3way |
anli_r3 |
ANLI R3 | NLI | 1,200 | 3way |
wanli |
WANLI | NLI | 5,000 | 3way |
lingnli |
LingNLI | NLI | 7,361 | 3way |
glue_diagnostic |
GLUE diag | NLI | 1,100 | 3way |
sick |
SICK | NLI | 4,906 | 3way |
scitail |
SciTail | NLI | 2,126 | 2way |
hans |
HANS | Reasoning | 15,000 | 2way |
med |
MED | Reasoning | 5,380 | 2way |
nan_nli |
NaN-NLI | Reasoning | 258 | 3way |
scone_test |
ScoNe | Reasoning | 1,000 | 3way |
nmonli_test |
NMoNLI | Reasoning | 200 | 3way |
conjnli |
ConjNLI | Reasoning | 623 | 3way |
nope_adv |
NOPE adv | Reasoning | 345 | 3way |
fol |
FOL | Reasoning | 10,277 | 3way |
control |
ConTRoL | Reasoning | 800 | 3way |
vitaminc |
VitaminC | Grounding | 10,000 | 3way |
dialfact |
DialFact | Grounding | 1,499 | 3way |
frank_sent |
FRANK | Grounding | 3,176 | 2way |
ragtruth |
RAGTruth | Grounding | 1,497 | 3way |
seahorse |
SEAHORSE | Grounding | 800 | 2way |
wice |
WiCE | Grounding | 958 | 2way |
docnli |
DocNLI | Grounding | 1,000 | 2way |
pubhealth |
PUBHEALTH | Grounding | 600 | 3way |
healthver |
HealthVer | Grounding | 230 | 3way |
scifact |
SciFact | Grounding | 300 | 3way |
findver |
FinDVer | Grounding | 500 | 2way |
nli4ct |
NLI4CT | Domain | 199 | 2way |
sdoh_nli |
SDOH-NLI | Domain | 990 | 2way |
biosift |
BioSift | Domain | 1,000 | 2way |
bionli |
BioNLI | Domain | 600 | 2way |
scientsbank |
SciEntsBank | Domain | 971 | 3way |
beetle |
Beetle | Domain | 819 | 3way |
legal_domain |
Legal | Domain | 1,163 | mixed |
contractnli_a |
ContractNLI A | Domain | 1,707 | 3way |
contractnli_b |
ContractNLI B | Domain | 2,074 | 3way |
vast |
VAST | Domain | 1,500 | 3way |
zs_meta |
Zero-shot meta | Domain | 5,400 | 2way |
fin_topics |
Fin topics | Domain | 600 | 2way |
crossingguard |
CrossingGuard | Domain | 5,397 | 3way |
- Downloads last month
- 106