Avoiding use of “girl” when you aren’t asking for a feminine child is good advice for any model that isn't specifically trained on a specialized prompting vocabulary (e.g., danbooru tags) where “girl” has a different meaning.
But I haven't seen anything reliable about H3 being particularly special in ther regard.
are these still 2026 problems? a simple vector index would associate girl as a synonym for adult woman as is used by both men and women in the English lexicon
even a Chinese model that thinks its Claude when asked would have inherited this association
I don't have hardware to runs this, so no, I didn't.
One thing that I had experience with, that led me to believe this might be true: it seems one of the earlier llama models was over-tuned to resist generating CSAM. Once I tried a somewhat sensitive prompt containing the word "girl" in it. Llama only ever generated refusals for this prompt, citing I was prompting for CSAM. GPTs and Claudes of that era had no issues with the same prompt.
Comments
One thing I read on this topic on reddit is never EVER use the word "girl" when prompting H3. So CSAM probably.
Avoiding use of “girl” when you aren’t asking for a feminine child is good advice for any model that isn't specifically trained on a specialized prompting vocabulary (e.g., danbooru tags) where “girl” has a different meaning.
But I haven't seen anything reliable about H3 being particularly special in ther regard.
are these still 2026 problems? a simple vector index would associate girl as a synonym for adult woman as is used by both men and women in the English lexicon
even a Chinese model that thinks its Claude when asked would have inherited this association
Did you try this yourself? Of course you wouldn't, because not wanting to produce SCAM sorry I meant CSAM.
And no, including the word "girl" in H3 does not lead to CSAM in any way, shape or form, but it's a great example how FUD quickly spreads.
I don't have hardware to runs this, so no, I didn't.
One thing that I had experience with, that led me to believe this might be true: it seems one of the earlier llama models was over-tuned to resist generating CSAM. Once I tried a somewhat sensitive prompt containing the word "girl" in it. Llama only ever generated refusals for this prompt, citing I was prompting for CSAM. GPTs and Claudes of that era had no issues with the same prompt.