notesum.ai
Published at November 25SAGEval: The frontiers of Satisfactory Agent based NLG Evaluation for reference-free open-ended text
cs.CL
cs.MA
Released Date: November 25, 2024
Authors: Reshmi Ghosh1, Tianyi Yao1, Lizzy Chen1, Sadid Hasan1, Tianwei Chen1, Dario Bernal1, Huitian Jiao1, H M Sajjad Hossain1
Aff.: 1Microsoft

| Scoring Criteria | Neg | Pos | Total | Definition |
| Accuracy | 41 | 0 | 41 | 2 |
| Semantic | ||||
| Diversity | 31 | 0 | 31 | 12 |
| Cohesion | 15 | 0 | 15 | 2 |
| Relevancy | 15 | 0 | 15 | 4 |
| Audience | ||||
| Understandability | 9 | 1 | 10 | 11 |
| Audience | ||||
| Engagement | 76 | 5 | 81 | 11 |
| Fairness | 5 | 1 | 6 | 7 |
| Sentiment | 1 | 10 | 11 | 32 |