3 ms·
They do not train newer models on output from older versions. In fact, they deliberately try to filter out LLM-generated text from their scraped internet datase
by Legend2440 2y ago
They do not train newer models on output from older versions. In fact, they deliberately try to filter out LLM-generated text from their scraped internet datasets.