BTZSC - NLI: leaderboard
BTZSC NLI cross-encoder slice measures zero-shot text classification macro-F1 for entailment models.
Metric: Macro-F1 (%). Source: huggingface.co. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | deberta-v3-large-nli-triplet | 59.58 | |
| 2 | deberta-v3-large-nli | 59.11 | |
| 3 | modernbert-large-nli | 55.17 | |
| 4 | deberta-v3-base-nli | 55.04 | |
| 5 | modernbert-large-nli-triplet | 54.87 | |
| 6 | modernbert-base-nli | 53.43 | |
| 7 | bert-large-uncased-nli | 53.39 | |
| 8 | bert-large-uncased-nli-triplet | 52.46 | |
| 9 | bart-large-mnli | 50.79 | |
| 10 | nli-roberta-base | 48.85 |
Interactive version: theaggregate.ai/benchmark?slug=btzsc-nli · How It Works · Data refreshed daily, snapshot 2026-10-20.