Aman Sanger: evaluation

Transcript context

…Yeah. Right. Well, you'll be able to fine tune OpenAI soon enough. We'll see. So we'll just get your general takes on LLM topics, just kind of quick fire, and then we'll go to lightning round. So human eval, that is the predominant way to benchmark code models because OpenAI benchmarks code models that way. There's some issues with it. Yeah. With open source models and even probably with some closed source models, it's unclear how much of it has actually leaked into the train set, right? So there's a recent model, New Hope, which it looked like they had some leakage, which is why it had really, really good performance. But I think there was an interesting approach taken by Palm too, where I think this is actually possible for someone to do right now. I've been meaning to do it at some point, but there's this paper called Babel code and they have a library which I think literally translates human eval into all other languages. And I think that would be a really good test because the other issues, a lot of the models that perform really well on human eval are pure Python, right? And that doesn't really give you a sense of if it's a good coding model overall. So yeah, I think at some point it would be really helpful if just someone did the work and ran the Babel code engine and translated human eval into all these other languages and then was able to run it. I think that would probably be a better benchmark, but still. I think if the original human eval problems leaked, I suspect it would also be helpful for solving the problems translated into other languages. But the issue is it's just so easy to run and anything else is probably going to be quite painful. Right? Well, it'd be better if there was a sandbox to run it. So agent cloud, hashtag. Hot take on training. Yeah.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.