Text Classification on 20NG
89.5AccuracyRoBerta-GCN
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| RoBerta-GCN2026.04 | 89.5 | — | — | |
| Bert-GCN2026.04 | 89.3 | — | — | |
| TensorGCN2026.04 | 87.74 | — | — | |
| TextGCN2026.04 | 86.34 | — | — | |
| BERT-base2026.04 | 85.7 | — | — | |
| G-LossBackbone=RoBERTa-large2026.04 | 85.33 | — | — | |
| SCL + CEBackbone=RoBERTa-large2026.04 | 85.31 | 84.9 | — | |
| G-Loss-O + CEBackbone=RoBERTa-large2026.04 | 85.19 | 84.23 | — | |
| G-LossBackbone=BERT-base2026.04 | 85.13 | — | — | |
| G-Loss-SQRT + CEBackbone=RoBERTa-large2026.04 | 85.06 | 84.61 | — | |
| CEBackbone=RoBERTa-large2026.04 | 84.89 | 84.28 | — | |
| GLoss-O + CEBackbone=BERT-base-uncased2026.04 | 84.71 | 84.17 | — | |
| SCL + CEBackbone=BERT-base-uncased2026.04 | 84.39 | 83.99 | — | |
| GLoss-SQRT + CEBackbone=BERT-base-uncased2026.04 | 84.01 | 83.52 | — | |
| CEBackbone=BERT-base-uncased2026.04 | 83.98 | 83.49 | — | |
| RoBERTa-large2026.04 | 83.8 | — | — | |
| GLoss-O + CEBackbone=DistilBERT-base2026.04 | 83.64 | 83.23 | — | |
| GLoss-SQRT + CEBackbone=DistilBERT-base2026.04 | 83.61 | 83.16 | — | |
| SCL + CEBackbone=DistilBERT-base2026.04 | 83.49 | 83.18 | — | |
| CEBackbone=DistilBERT-base2026.04 | 83.15 | 82.69 | — | |
| BertGCNDescription=Co-trained BERT with TextGCN2026.04 | — | — | 89.5 | |
| ChatGPTEvaluation Protocol=Zero-shot2026.04 | — | — | 58.7 | |
| ChatGPTEvaluation Protocol=Few-shot, k=22026.04 | — | — | 58.44 | |
| Linear ProbingEmbedding Type=Pre-trained SBERT (P)2026.04 | — | — | 79.06 | |
| Linear ProbingEmbedding Type=Fine-tuned SBERT (F)2026.04 | — | — | 86.9 | |
| LREmbedding Type=Pre-trained SBERT (P)2026.04 | — | — | 79.06 | |
| LREmbedding Type=Fine-tuned SBERT (F)2026.04 | — | — | 87.17 | |
| ModTGCNEmbedding Type=Pre-trained SBERT (P)2026.04 | — | — | 90.6 | |
| ModTGCNEmbedding Type=Fine-tuned SBERT (F)2026.04 | — | — | 91.14 | |
| ModTGCNEmbedding Type=Fine-tuned SBERT (F), Warmup=true2026.04 | — | — | 89.91 | |
| ModTGCN(F)Embedding Setting=Fine-tuned2026.04 | — | — | 91.14 | |
| ModTGCN(P)Embedding Setting=Pre-trained2026.04 | — | — | 90.6 | |
| TensorGCN2026.04 | — | — | 87.74 | |
| TextGCN2026.04 | — | — | 86.34 | |
| U-TextGCN2026.04 | — | — | 86.07 | |
| WCTextGCN2026.04 | — | — | 87.21 |