On: Learning to Reason with LLMs

[apsec112]

If there's a way to do something like this with Llama I'd love to hear about it (not being sarcastic)

[catchnear4321 → apsec112]

nurture the model have patience and a couple bash scripts

[apsec112 → catchnear4321]

But what does that mean? I can't do "pip install nurture" or "pip install patience". I can generate a bunch of answers and take the consensus, but we've been able to do that for years. I can do fine-tuning or DPO, but on what?

[catchnear4321 → apsec112]

you want instructions on how to compete with OpenAI?

go play more, your priorities and focus on it being work are making you think this to be harder than it is, and the models can even tell you this.

you don’t have to like the answer, but take it seriously, and you might come back and like it quite a bit.

you have to have patience because you likely wont have scale - but it is not just patience with the response time.