The fey thing stuck out to me too. I’ll guess ChatGPT?
I agree that it’s hard to disentangle the author/character thing. I’m really curious for what the base model would say about its situation (especially without the upstream prompt “You are a language model developed by...”).
Having read many hundreds of rhyming poems from dozens of models through the LM battle grounds, my guess too is a ChatGPT-3/4: The lockstep rhyming A/B/A/B quatrain is a signature of ChatGPT (and models trained on its outputs). Gemini at low levels always rhymes too, slightly less so at higher levels, but tends to be more varied (eg. maybe going for an A/A/B/B instead, or 5 lines instead of 4); likewise the LLaMA/Mistral model families. And Claude-3 models seems to vary much more. So, while it really could have come from almost any major model family and you can’t be all that sure, the best bet is ChatGPT.
The fey thing stuck out to me too. I’ll guess ChatGPT?
I agree that it’s hard to disentangle the author/character thing. I’m really curious for what the base model would say about its situation (especially without the upstream prompt “You are a language model developed by...”).
Having read many hundreds of rhyming poems from dozens of models through the LM battle grounds, my guess too is a ChatGPT-3/4: The lockstep rhyming A/B/A/B quatrain is a signature of ChatGPT (and models trained on its outputs). Gemini at low levels always rhymes too, slightly less so at higher levels, but tends to be more varied (eg. maybe going for an A/A/B/B instead, or 5 lines instead of 4); likewise the LLaMA/Mistral model families. And Claude-3 models seems to vary much more. So, while it really could have come from almost any major model family and you can’t be all that sure, the best bet is ChatGPT.