Feedback
Funny :D
Expected: Male, genitals, duo, clothing and outside tags should not have emerged.
Why? They are plain wrong in this context and very unexpected.
Hypothesis: I assume the word "riding" is highly correlated to pornographic topics, so maybe that's why those tags were triggered
The training data of these two experimental models are very limited. If there are some interesting use cases (and if I have time), maybe I'll retrain them to be more useful :D
Thanks :-) I think Prompt to Tag and Tag to Prompt is very useful.
Use-cases:
- Prompt2Tag: Tags for dataset classification
- Prompt2Tag: Tags for clustering
- Prompt2Tag: Tags as pre-condition for a more granular sofisticated search algorythm
- Prompt2Tag: Use tags as form of annotation
- Tag2PRompt: Dataset creation by using tags as basis for seeds.
I am sure those two models (if they work great) would also fit well into an agent framework. Maybe even make them fit into Ant-Think syntax or something similar.
I am sure there are more use-cases. Those are just a few I cam up with on the spot.