>>106127755
Llama-3.3-Nemotron-Super-49B-v1.5 provided the text and then I asked Qwen to do the styling. AND IT LOST THE GOOD BITS. Or maybe not, maybe good guy Qwen made it more tight and concise.
>other 4chan-specific terms (loli, bread meaning post/points)
>>106127766
It's Nvidia, they use crappy datasets for finetuning, making models worse for rp. Their only good Nemotron was 4-340B they trained from scratch.