I wasnt asking so much about recording the presentations in Sydney, but about prerecorded videos like FOCS 2024 and like STOC has done for many years already

ah yes, i think that works

When you succeed on the disadvantaged roll, you don't notice that you rolled wrongly. Hence you cannot simply reroll straight: you would effectively have a too-large success probability.

Surely there is an easy probability thing you can do in this situation?

I really should have considered timezones before tweeting at Australia

this is one of my favorite papers of 2025

oh yes this. it's hard to relate these works, since something like corollary 1.1 is false in any situation where simplex has polynomial running time

your question is an interesting one, and i hope to one day see more research on it. but right now we dont know much

this doesn't ring any bells for me

i use doi2bib.org. paste in the doi, copy out the bibtex entry.

i assume the value is not so much the amount of change, but the relation between the LLM output given and the changes made in response

this. also, authors are asked to give feedback on the LLM output. that's highly skilled labeling work that the authors are not being paid for

automated pre-submission feedback
acm-stoc.org/stoc2026/sto...

wait what? thats so cool, and so very early!

i feel second-hand embarrassment if you do this. words mean things

The online reception has been heartwarming. Great to see people being as enthusiastic about our result as we the authors feel 🥰