Performance prediction on Sentiment temporal (out-of-domain)
0.834ROC AUCCosine distance (fine-tuned)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Cosine distance (fine-tuned)Model dependency=model-dependent, Embedding status=fine-tuned2023.05 | 0.834 | 236.7 | |
| Vocabulary, structural, semantic driftModel dependency=model-agnostic2023.05 | 0.596 | 84.8 | |
| Semantic driftModel dependency=model-agnostic2023.05 | 0.586 | 110.4 | |
| Structural driftModel dependency=model-agnostic2023.05 | 0.581 | 146.1 | |
| Vocabulary driftModel dependency=model-agnostic2023.05 | 0.571 | 105.8 | |
| Combined prev. model-agnosticModel dependency=model-agnostic2023.05 | 0.562 | 142.1 | |
| Cosine distance (pre-trained)Model dependency=model-agnostic, Embedding status=pre-trained2023.05 | 0.559 | 91.8 | |
| Token frequency cross-entropyModel dependency=model-agnostic2023.05 | 0.557 | 97.3 | |
| Token frequency JS-divModel dependency=model-agnostic2023.05 | 0.528 | 106.2 | |
| Baseline (no-performance drop)Model dependency=model-agnostic2023.05 | 0.5 | 100 |