Multi-label Value Classification on Gemini-annotated N=4,000 (test)
0.752Self-direction ScoreXLM-RoBERTa-large
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| XLM-RoBERTa-largeBackbone=XLM-RoBERTa-large, Fine-tuning protocol=last two layers, Pooling strategy=mean pooling, Classification head=linear multi-label classification head, Maximum sequence length=512, Loss function=binary cross-entropy (BCE), Batch size=16, Optimizer=AdamW, Learning rate=2e-5, Training epochs=62026.06 | 0.752 | 0.633 | 0.726 | 0.794 | 0.473 | 0.686 | 0.429 | 0.755 | 0.93 | 0.702 |