Authorship segmentation on MAS (entire corpus)
45.7SBDA @ 0.3RoBERTa + ModernBERT + CRF*
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| RoBERTa + ModernBERT + CRF*Type=Proposed Approach, Info-Mask=true, backbone=RoBERTa + ModernBERT2025.12 | 45.7 | 44.4 | 42.1 | 37.2 | 41.4 | 40.2 | 38.7 | 33.7 | |
| ModernBERT + BiGRU + CRF*Type=Proposed Approach, Info-Mask=true, backbone=ModernBERT, recurrent_layer=BiGRU2025.12 | 41.9 | 41.2 | 39.8 | 36.3 | 40.2 | 39.5 | 38.2 | 34.8 | |
| RoBERTa + BiGRU + CRF*Type=Proposed Approach, Info-Mask=true, backbone=RoBERTa, recurrent_layer=BiGRU2025.12 | 41.6 | 40.8 | 38.3 | 35.2 | 32.5 | 31.9 | 30.3 | 27.5 | |
| ModernBERT + CRF*Type=Proposed Approach, Info-Mask=true, backbone=ModernBERT2025.12 | 41.3 | 40.7 | 39.3 | 35.8 | 38.6 | 38 | 36.8 | 33.5 | |
| DeBERTa + BiGRU + CRF*Type=Proposed Approach, Info-Mask=true, backbone=DeBERTa, recurrent_layer=BiGRU2025.12 | 39.8 | 39.2 | 37.9 | 35.2 | 37.1 | 36.6 | 35.3 | 32.8 | |
| DeBERTa + CRF*Type=Proposed Approach, Info-Mask=true, backbone=DeBERTa2025.12 | 38.7 | 38.1 | 36.9 | 34.7 | 38.2 | 37.8 | 36.9 | 34.2 | |
| RoBERTa + CRF*Type=Proposed Approach, Info-Mask=true, backbone=RoBERTa2025.12 | 32 | 20 | 8 | 2 | 0.5 | 0.3 | 0.1 | 0 | |
| SeqXGPTType=Prior Baselines, Info-Mask=false2025.12 | 31.5 | 27.3 | 22.8 | 15 | 30.1 | 26.4 | 21.9 | 14.8 | |
| SpanBERTType=Prior Baselines, Info-Mask=false2025.12 | 29.4 | 25.8 | 19 | 11.2 | 28.1 | 24.9 | 18.3 | 10.9 | |
| BinocularsType=Zero-shot Span-Scoring, Info-Mask=false2025.12 | 28.5 | 23.5 | 7.9 | 2.2 | 28.5 | 23.5 | 7.9 | 2.2 | |
| BiLSTM+CRFType=Prior Baselines, Info-Mask=false2025.12 | 26.1 | 21.5 | 15.3 | 8.1 | 25.3 | 20.8 | 14.9 | 7.8 | |
| FastDetectGPT (gpt-neo-2.7b)Type=Zero-shot Span-Scoring, Info-Mask=false, backbone=gpt-neo-2.7b2025.12 | 25.9 | 20.8 | 6.4 | 1.5 | 25.9 | 20.8 | 6.4 | 1.5 | |
| FastDetectGPT (falcon-7b-instruct)Type=Zero-shot Span-Scoring, Info-Mask=false, backbone=falcon-7b-instruct2025.12 | 25 | 21.4 | 8.4 | 1.9 | 25 | 21.4 | 8.4 | 1.9 | |
| GPTZeroType=Zero-shot Span-Scoring, Info-Mask=false2025.12 | 22.3 | 13.7 | 2.5 | 0.8 | 28.9 | 14.2 | 8.5 | 1.2 | |
| Glimpse (davinci-002)Type=Zero-shot Span-Scoring, Info-Mask=false, config=davinci-0022025.12 | 21.4 | 16.8 | 4.4 | 1.1 | 21.4 | 16.8 | 4.4 | 1.1 | |
| Glimpse (baggage-002)Type=Zero-shot Span-Scoring, Info-Mask=false, config=baggage-0022025.12 | 20.8 | 16.1 | 4 | 0.9 | 20.8 | 16.1 | 4 | 0.9 | |
| logp(x)Type=Statistical Based, Info-Mask=false2025.12 | 15.3 | 10.7 | 1.3 | 0.5 | 15.3 | 10.7 | 1.3 | 0.5 | |
| entropyType=Statistical Based, Info-Mask=false2025.12 | 15.1 | 10.5 | 1.1 | 0.3 | 15.1 | 10.5 | 1.1 | 0.3 |