3 ms·
and then they use the output of chatGPT to train their open models
by seydor 2y ago
and then they use the output of chatGPT to train their open models
- 123yawaworht456 2y agowhich is a pity, because the models and finetunes tainted with even a minuscule amount of GPT slop are affected very badly. you can easily tell the difference between llama finetunes with or without synthetic datasets.