| system | n | chrf | edit_sim | rougeL | exact_match | number_recall | placeholders_exact |
|---|---|---|---|---|---|---|---|
| B0_identity | 2574 | 63.999 [61.898, 65.824] | 0.649 [0.631, 0.664] | 0.654 [0.635, 0.671] | 0.000 [0.000, 0.000] | 1.000 [1.000, 1.000] | 1.000 [1.000, 1.000] |
| B1_generic_paraphrase | 2574 | 61.950 [60.145, 63.494] | 0.533 [0.520, 0.546] | 0.570 [0.554, 0.585] | 0.000 [0.000, 0.000] | 1.000 [1.000, 1.000] | 1.000 [1.000, 1.000] |
| htr-m1-small | 2574 | 63.333 [61.317, 65.084] | 0.640 [0.623, 0.655] | 0.645 [0.628, 0.661] | 0.000 [0.000, 0.000] | 0.992 [0.984, 0.998] | 0.999 [0.998, 1.000] |
| htr-m1-tiny-s2 | 2574 | 63.455 [61.374, 65.272] | 0.644 [0.627, 0.660] | 0.649 [0.632, 0.665] | 0.000 [0.000, 0.000] | 0.994 [0.989, 0.999] | 0.997 [0.993, 1.000] |
| htr-m1-tiny-s3 | 2574 | 63.475 [61.408, 65.291] | 0.643 [0.627, 0.659] | 0.648 [0.631, 0.664] | 0.000 [0.000, 0.000] | 0.994 [0.988, 0.998] | 0.999 [0.998, 1.000] |
| htr-m1-tiny | 2574 | 63.458 [61.433, 65.238] | 0.644 [0.627, 0.659] | 0.648 [0.632, 0.664] | 0.000 [0.000, 0.000] | 0.993 [0.985, 0.998] | 0.999 [0.998, 1.000] |
| htr-m2-tiny | 2574 | 63.264 [61.138, 65.116] | 0.641 [0.624, 0.657] | 0.647 [0.629, 0.663] | 0.000 [0.000, 0.000] | 0.992 [0.985, 0.998] | 0.997 [0.990, 1.000] |
| htr-m3light-tiny | 2574 | 63.143 [61.077, 64.951] | 0.641 [0.624, 0.656] | 0.646 [0.628, 0.662] | 0.000 [0.000, 0.000] | 0.993 [0.986, 0.998] | 0.998 [0.994, 1.000] |
| htr-m3tokenonly-tiny-s2 | 2574 | 63.425 [61.339, 65.254] | 0.644 [0.627, 0.660] | 0.650 [0.632, 0.666] | 0.000 [0.000, 0.000] | 0.996 [0.993, 0.998] | 0.999 [0.996, 1.000] |
| htr-m3tokenonly-tiny | 2574 | 63.396 [61.287, 65.254] | 0.645 [0.628, 0.661] | 0.650 [0.632, 0.666] | 0.000 [0.000, 0.000] | 0.995 [0.989, 0.998] | 0.998 [0.994, 1.000] |
Stylometric W1 (mean over features, standardised; lower = closer to originals):
- rewrite vs original: 0.201
- B0_identity vs original: 0.201 · probe P(original) mean 0.41
- B1_generic_paraphrase vs original: 0.227 · probe P(original) mean 0.47
- htr-m1-small vs original: 0.136 · probe P(original) mean 0.51
- htr-m1-tiny-s2 vs original: 0.143 · probe P(original) mean 0.50
- htr-m1-tiny-s3 vs original: 0.137 · probe P(original) mean 0.51
- htr-m1-tiny vs original: 0.140 · probe P(original) mean 0.51
- htr-m2-tiny vs original: 0.135 · probe P(original) mean 0.52
- htr-m3light-tiny vs original: 0.137 · probe P(original) mean 0.53
- htr-m3tokenonly-tiny-s2 vs original: 0.140 · probe P(original) mean 0.50
- htr-m3tokenonly-tiny vs original: 0.147 · probe P(original) mean 0.49