News for hackers. Stuff that matters.

A comment on Building LLMs from the Ground Up: A 3-Hour Coding Workshop. Parent comment

Re:Building LLMs from the Ground Up: A 3-Hour Coding Workshop by rahimnathwani on Saturday August 31, 2024 @10:59PM (#41413889) 

  how Llama and OpenAI could be cleaning and structuring their training data
If you're interested in this, there are several sections in the Llama paper you will likely enjoy:

https://ai.meta.com/research/publications/the-llama-3-herd-o...

[ Parent | on HN ]
Re:Building LLMs from the Ground Up: A 3-Hour Coding Workshop by kbrkbr on Sunday September 01, 2024 @10:22AM (#41417188)
But isn't it the beauty of llm's that they need comparably little preparation (unstructured text as input) and pick the features on their own so to say?

edit: grammar

[ Parent | on HN ]
Re:Building LLMs from the Ground Up: A 3-Hour Coding Workshop by aDyslecticCrow on Monday September 02, 2024 @10:19AM (#41425582)
Yes, if you want an LLM that doesn't listen to instructions and just endlessly babbles about anything and everything.

What turned GPT into chatGPT was a lot of structured training with human feedback.

[ Parent | on HN ]
Re:Building LLMs from the Ground Up: A 3-Hour Coding Workshop by rahimnathwani on Tuesday September 03, 2024 @11:47PM (#41441714)
Exactly. Section 4.3.7 briefly explains how they trained the model to better follow instructions ('steerability').
[ Parent | on HN ]