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 months

No recordings in the last 12 months.Older appearances are listed below; set an alert to hear about the next one.

newest first · ▶ plays the moment

But essentially it takes steps to learn that this represent and it it's it's Bellman updates.

It backdrops from this final reward is how the valve value model learned.

So it takes time to get

to the earlier tokens, which is what the warm up is doing.

like low value and then the actual answer would have high value.

Or the answer is might have high value because it knows that the answer normally follows it.

Or something like this.

I think it's it's probably not easy for us to build intuitions on this because the the feature space is so big.

It's like like how are you gonna build intuitions over what like in o in O one's case it's like w is wait a high value token?

But if it's like too high value, it would only just be like wait, wait, wait, and never answer.

So there's definitely some like weird things that I don't know how clear it is to make

to get intuitions out of it.

So it's like not just the token.

It's like the token in context.

A good direction.

Inference time compute is very related to R L.

Where it's like

Having a good model from RL that can attribute value is very useful to spending more on inference and

We'll see all of these things continue to proliferate.

And I put entropics more in the

Showing 481–500 of 1,013 · page 25 of 51