Der Faschismusexperte Jason Stanley stellt Deutschland in diesem Interview (gegen Ende des langen Texts) ein schlechtes Zeugnis für seinen Kampf gegen den Neofaschismus aus. Lesenswert!
www.fr.de/kultur/gesel...

LLM PR is "impressive" in that it picks up a lot of the issues that human reviewers do. Or at least, the issues that humans do based on a quick read. And that's the problem with both human and LLM reviews. They respond mostly to surface level. In humans because of time pressure, in LLMs by design.

I have to try more CSD stuff as well - always seemed a bit arbitrary to me 🙈

I'm unfortunately bogged down with too many projects to push the decisionmaking angle strongly right now, so it will take time :)

Tricky indeed - we are working on getting RIDE to combine nicely with unfold, maybe that's a good venue. and Cameron-Hassals time-dilation could also be an option for modelling. All with more realistic simulations of SSMs via UnfoldSim

(one more thing, could you also upload the continuous EEG data, not only the epoched? that would be super great :-))

uh and this looks like amazing work with a super great experimental design, cool analyses and nice datasharing. Thanks a ton!

thanks for tagging!

Quick q: Given the C-response is what supposedly is the actual decision making process - why not analyse that directly? Don't you else have an overlap-biased c-component overlap with R?

In other words, why not look at C directly?

MRI is one of the coolest medical technology!

Dont grasp the relevance of your example unfortunately. Do you describe an implicit small SE over models? Maybe this could help me: if a previous model has e.g. 68.3% accuracy, and your model has 68.5% - how do you decide this is relevant and not just a lucky init, split, dataset sampling bias etc.?

How do you do inferential statistics yourself? There should be your answer.

What's the unit of observation & how do you model dependencies?

I always wondered when I've seen those bolded tables claiming superiority ¯\_ (ツ)_/¯

@olaf.dimigen.de