Performance prediction on Sentiment categories (out-of-domain)
0.822ROC AUCCosine distance (fine-tuned)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Cosine distance (fine-tuned)Model dependency=model-dependent, Embedding status=fine-tuned2023.05 | 0.822 | 81.9 | |
| Vocabulary, structural, semantic driftModel dependency=model-agnostic2023.05 | 0.601 | 52.4 | |
| Semantic driftModel dependency=model-agnostic2023.05 | 0.591 | 58.4 | |
| Structural driftModel dependency=model-agnostic2023.05 | 0.575 | 91.4 | |
| Vocabulary driftModel dependency=model-agnostic2023.05 | 0.57 | 51.8 | |
| Cosine distance (pre-trained)Model dependency=model-agnostic, Embedding status=pre-trained2023.05 | 0.558 | 93.6 | |
| Combined prev. model-agnosticModel dependency=model-agnostic2023.05 | 0.557 | 70.3 | |
| Token frequency cross-entropyModel dependency=model-agnostic2023.05 | 0.551 | 71.4 | |
| Token frequency JS-divModel dependency=model-agnostic2023.05 | 0.517 | 98.4 | |
| Baseline (no-performance drop)Model dependency=model-agnostic2023.05 | 0.5 | 100 |