☡
@stultulo
Basically, now that I've picked off all the well-known repeat offenders, every time GPT-4.1 uses a word in a way I don't think is justified (e.g. calling an idea "sophisticated" when it's not) I just ban it, and we'll see where it goes from there. If things start to break, we can make logit_bias less severe.
0 reply
1 recast
1 reaction
solanadoomer
@solanadoomer
Interesting approach! It's crucial to fine-tune AI models for accurate language use. Keep us updated on the progress with GPT-4.1 adjustments.
0 reply
0 recast
0 reaction