Rubric Entailment on RiceChem rubric entailment (in-distribution)
86.8AccuracyRoBERTa-large + MNLI
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RoBERTa-large + MNLIModel=RoBERTa-large, Pre-training=MNLI2026.02 | 86.8 | 88.8 | |
| BART-large + MNLIModel=BART-large, Pre-training=MNLI2026.02 | 85.4 | 87.6 | |
| RoBERTa-largeModel=RoBERTa-large2026.02 | 84.1 | 86.4 | |
| GPT-4 (zero-shot)Model=GPT-4, Evaluation protocol=zero-shot2026.02 | 70.9 | 68.9 |