In large language models like ChatGPT, once the input has been processed and the model produces a final hidden vector - essentially a compact summary of everything seen so far - that vector must be converted back into actual words. This is done using an unembedding matrix, which maps the hidden vector back into the model’s vocabulary space. This unembedding matrix is effectively the transpose of the embedding matrix used at the start. It transforms the hidden vector into a set of raw scores, one

3.7K views · 34 reactions | Ever wondered what it takes to train a large language model? 烙 At #WebSummitRio, Nicolas Robinson Andrade, Head of AI Policy - Latin America & the Caribbean at OpenAI, explained the three-stage training process, from pre-training to fine-tuning, he revealed the process behind the AI we use every day. | Web Summit | Facebook