Nathan Lambert
speaker
1,013 appearances
1 recordings
1 series
first heard Nov 2024
last heard Nov 2024
Nathan Lambert’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsNo recordings in the last 12 months.Older appearances are listed below; set an alert to hear about the next one.
Appearances
Inference time compute.
Thing but there's very similar
intuitions.
Like if you sample more within a bat if you sample more completions from a prompt.
in RL, you're essentially like exploring more to potentially find a high value completion.
And if you're doing inference time compute, a lot of it comes down to how do you sample from the language model and encourage it and like reweight or interrupt it or change the sampling to find the right completion.
And that's where like Monte Collar tree search type things come in is like
search over reasoning steps or all of these like Q star or whatever type of things you want to go down that direction.
It is related, but I think
It's you can draw a boundary between training and inference and it's it's okay.
back and try again.
Well, I would say that I do think that something like this is what
Like.
O one could be doing.
But just like we see how easy it is for RL to go off the rails.
Like if you run RL for a long time on something like IFAVAL, which is these constraints, it literally does exacolate the constraints.
It's like, please include this word in your answer and make your response a hundred words or longer.
It literally just will print the word a hundred times and then stop.
It's like it's like stupid stuff like this.
And it's like, well, maybe if you can do RL.
Showing 501–520 of 1,013 · page 26 of 51
← Previous
Next →