ModernBERT-large + CRF — Exp F
Fine-tuned answerdotai/ModernBERT-large with CRF decoding.
Architecture
ModernBERT-large -> Dropout(0.3) -> Linear(hidden->5) -> CRF (Viterbi decoding)
Comparison
| Label | Paper | Exp A | Exp C | Exp E | Exp F (CRF) |
|---|---|---|---|---|---|
| FP | 1.000 | 0.9944 | 0.9944 | 0.9915 | 0.9944 |
| RP | 0.690 | 0.8022 | 0.8964 | 0.8802 | 0.8620 |
| RV | 0.400 | 0.3145 | 0.4974 | 0.4884 | 0.2857 |
| PW | 0.830 | 0.8879 | 0.9451 | 0.9409 | 0.9345 |
| Macro | 0.730 | 0.7497 | 0.8333 | 0.8253 | 0.7691 |
| Label | P | R | F1 | Support |
|---|---|---|---|---|
| O | 0.9793 | 0.9951 | 0.9871 | 3704 |
| FP | 0.9888 | 1.0000 | 0.9944 | 176 |
| RP | 0.8453 | 0.8793 | 0.8620 | 174 |
| RV | 0.4500 | 0.2093 | 0.2857 | 86 |
| PW | 0.9857 | 0.8884 | 0.9345 | 233 |
- Downloads last month
- 26