Sebastian Raschka

speaker

1,024 appearances 1 recordings 1 series first heard Feb 2026 last heard 1 Feb

Sebastian Raschka’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

Recordings per month over the last 12 months — 1 in all, peaking in Feb 2026 with 1.

newest first · ▶ plays the moment

So I think the interesting, beautiful thing here is that you ask the LMA, let's say a math question, and then you know the correct answer.

And you let the LLM, like you said, figure it out.

But how it does that, I mean, you don't really constrain it much.

There are some constraints you can add, like use the same language, don't switch between Spanish and English.

But let's say you're pretty much hands-off.

You only give the question and the answer.

And then the LLM has to, you know, just the task to arrive at the right answer.

But the beautiful thing here is what happens in practice is that the LLM will do a step-by-step description.

Like, you know, like as a student or like as a...

yeah, mathematician, how you would derive the solution.

It will give you, or it will use those steps.

And that helps actually the model to improve its own accuracy.

And then, like you said, the inference scaling.

So inference scaling loosely means basically spending more compute during using the LM during inference.

And here the inference scaling is,

that the model would use more tokens.

And also, I think in the R1 paper, they showed the longer they train the model, the longer the responses are.

They grow over time.

They use more tokens.

So it becomes more expensive, becomes more expensive for simple tasks.

Showing 421–440 of 1,024 · page 22 of 52