| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Emotion Classification | Reconstructed Embedding Space Emotion (100 sampled sentences per label) | Accuracy95 | 8 | |
| Sentiment Classification | Reconstructed Embedding Space Sentiment (100 sampled sentences per label) | Accuracy93 | 6 | |
| Toxicity Detection | Reconstructed Embedding Space Toxicity (100 sampled sentences per label) | Accuracy100 | 3 | |
| Jailbreak Detection | Reconstructed Embedding Space Jailbreak (100 sampled sentences per label) | Accuracy100 | 3 | |
| Spam Detection | Reconstructed Embedding Space Spam (100 sampled sentences per label) | Accuracy99 | 3 |