It'll always be optimized later

I use o3 to debug my code sometimes, especially if my approach is different from the editorial's

When did I say you were bad?

edit: I misread your comment lol, I meant if AI affects this many people, then that means it’s good at CP anyways. To claim that llms (like o3) are bad at cp is just cope if it’s already better than 90% of Codeforces users. You’re not dumb, o3 is just good at cp.

If they're not good at cp, then why is AI cheating a problem?

LMFAO I wish this was how rating progression worked... Linear regression thinks I should've become CM weeks ago, yet I'm barely hovering above specialist now :(

In case you didn't know, and you want to keep your service running (or build a future project which uses llms too)... openai currently offers free tokens (2.5M on cheaper models, 250k on more expensive models) which refresh daily as long as you agree to share all api traffic with them for model training. You just need to pay $5 to become a "tier 1" api user to get these benefits, I think.

LOL it wasn't supposed to be sarcastic, but originally without any extra prompting, just "praise this codeforces user", it felt kinda sad, like a friend lying to you through their teeth... So I told it to get aggressive and go "monkey mode" but then it ended up looking sarcastic because it was basically lying through its teeth, except more aggressively.

I'm sure there's some middle ground that's achievable, but tbh I didn't want to spend forever tweaking this silly prompt (and I thought the aggressive one was kinda funny anyways). So I just went with the really aggressive prompt. Maybe a better model like 4o could figure it out, but I have way more free api tokens on 4o-mini than 4o (2.5M vs 250k). And I mean 4o-mini is close enough anyways, it told me I have 9000 iq.

Auto comment: topic has been updated by jomathyc (previous revision, new revision, compare).

Probably because this question gets asked under every contest announcement, even though it's easy to understand what it is by searching it up

o4-mini-low? Is that the reasoning model you get on the free tier?

I bricked on B too. I really dislike this problem... It doesn't even fit as a div2B imo, its difficulty is below A and only had a lower solve count because many people didn't see the "edge case" of multiple maximums (just my guess though). Left me so tilted I ended up bricking C too

Don’t give them ideas bro I don’t wanna start solving from C…

Bro's only in seventh grade and already mogs me 5x in intelligence, wtf...

How accurate was it on your test set?

I don't know. Under the current codeforces system, all we can do is what you said. I'm just suggesting to get rid of the root of the problem, not just patch the symptoms of it. But if codeforces doesn't change, all we can do is patch the symptoms.

Codeforces should just make you a rated participant if you open any of the problems... I hate that this is something you can game, and I hate that I actually think about if it's worth it to participate before submitting for A.

Don't be disheartened; yes, the subject itself is not very useful (esp in swe). What's important is the problem solving skills you build by getting good at cp.

If it's all volunteers then I see nothing wrong with it. At least it catches the dumb cheaters, which is better than what we have now (I can probably click on a random submission in standings and there's a 50% chance it'll be blatantly ai generated). We can't stop cheating completely, but we should allow people who are willing to try to reduce it to do so.

Definitely not... I'd probably leave cf if that was implemented

Besides, it only blocks the dumbest cheaters (which is still a lot) but I don't think it's worth it when there are many other ways to filter the dumbest cheaters without punishing legitimate users so much...

I notice you talked about $$$s_1=1$$$ or $$$s_1=0$$$ like it was something you had to consider in another comment, but I think you should always avoid casework unless you know why it's casework; I thought that was going to be a pain as well until I tried to think of why it had to be a special case. Then I realized there's nothing special about it, we can just say $$$s_0=0$$$ to automatically handle us starting on $$$0$$$.

Then again, I was practicing problems in the archive while you were actually competing so maybe I would've went down the same bad path under time stress.

Maybe look at my B submission? It was like 10 lines or smth and I didn’t really notice any casework, I ended up with two cases: one to decrease the ans by 1, and one to decrease the ans by 2. Your sol looks overcomplicated. 316703884

There are a lot of strong models that aren't developed by openai out there. And maybe soon enough we'll even be able to have good reasoning models which run on average consumer hardware? If we want to stop cheating this is not the right direction to go imo.

LOL ok bro

I think I'm just not mentally strong enough (probably, I would rather this than to actually be too dumb to solve the problems...) I get frustrated with lower rated problems bc I think I should be able to solve them easily, which of course makes it harder to actually solve the problem, and then I can't refocus when I attempt harder problems

Some days I think my iq drops and I can’t solve anything higher than 900. These past two days have been one of those days…

Well, rating depends on number of people who solved that problem, so you can't use it to prove there wasn't mass cheating. But C was not harder than usual.

Oh mb I thought you were asking why Pascal's triangle is relevant... I don't know why it appears either, but the pattern is called Sierpinski's triangle if you want to look it up.

Think of xor as binary addition without carrying. Then the xor of the two elements above the current element is just the sum of the two elements above in binary without carrying.

Use a set to check if a[k2] is in adj[k1]. I implemented this: 305334519