Lately I have been seeing many many blogs after the recent Div 3 contest (Codeforces Round 1109 (Div. 3)) regarding cheating and AI usage. While this post is not directly regarding that, one question has been constantly bothering me. That is — are Div 3 and Div 4 contests still relevant?
You can detect if someone copy-pastes an AI generated code. But let's say an user gives the problem statement to an LLM and asks it to break down the problem brick by brick. Or even worse, someone can literally give the AI one of his previous codes (that he wrote himself) and he can then ask to write the solution code in exactly same format (same spaces, same type of variables, human-like variable declaration, no advisory comments etc). This is no big deal if your prompting is not terrible. Nobody can ever detect this type of 'smart AI usage', which is obviously, strictly prohibited!
LLMs like Google Gemini and Anthropic Claude solve problems from Div 3 like it is nothing. There was a time when these LLMs could solve up to C-D, but now that bound is completely broken — it seems that an efficient AI can solve an entire contest today.
And even worse, I would like to share an experience of mine from the last Div 2 contest (Codeforces Round 1108 (Div. 2)). In this contest, I solved A, B and C. After trying a lot, I couldn't solve D, it kept getting wrong answer on pretest 3 (Here are my submissions during the contest: 382292963, 382297206). Finally after the contest, a little frustrated, I opened Gemini and gave the problem and my code. It gave me a 'corrected code' within a minute (382302565) and unexpectedly (or expectedly), it got accepted. Therefore LLMs are solving Problem D-s of Div 2 on first tries, now-a-days. God have mercy on lower divisions.
Edit 1: I seem to have failed to convey what I meant. The fun and excitement of any contest never changed, and it never will (hopefully). But is the rating still relevant? Personally I think the system needs to be adjusted a bit according to the present scenario, involving AI and all.
The relevancy of Codeforces contests is based on how much people enjoy and learn from them, not by how well people can cheat with AI
I completely agree with you. But fairness still matters.
Yeah, but at least the most important goal of codeforces contests rn is to gain some skill/enjoy problems or compete with your friends
I'd rather look at friends standings instead of public standings cuz its too disgusting.
If you participate in divs for the sake of rating only, then I suggest you don't participate anymore.
Don't get me wrong having a high rating is always very nice to have, but it's extremely irrelevant nowadays due to shitty people trying to prove who's the better prompt engineer, so I suggest you try and learn as much as possible, and enjoy the journey. Forget about cheaters, and you'll realise that Div 3s are extremely fun to participate in.
It is not why I myself participate, but fairness does matter a lot. And yes, Div 3-s are definitely extremely fun! Actually in my country, it is like a competition among college students about who has more rating. Everybody wants to be an expert — and people will do everything including cheating. That is my entire point, the 'fun' never changes and perhaps never will, but the rating is becoming more and more meaningless everyday. That's what I wanted to say. Nothing more.
Everyone is equally cooked, AIs can solve problems from every division. (XVIII) Although if you make it to the top of the standings, you are maybe a little safer, as (I think) admins prioritize the top of the leaderboard first.
Weird. Post-comp I tried using AI on D and it got the complete wrong answer: it didn't realize Bob's best strategy is to put an odd number on index 2 instead of 1.
Depends a lot on how you gave the input to the AI :D
I I gave it the problem and it just floped. I even gave it my own code that passed and it said that the judge had weak test cases :sob:
The point is that it doesn't matter, contests are made for humans by humans ( i hope ) , and many humans ( my self included ) like div3 and div4 because in those they can (usually) solve more problems and (usually) sharpen their observations since div3 and div4 do not require ( at least not in the inteded solutions ) crazy data structures nor crazy algorithms, just observations. LLM's if given enough power can already solve atcoder world finalist level problems, so based on what you are saying codeforces should just shut down tomorrow but that's not the case, yes, i hate being 1200 because many can jump over me by cheating, but in the end it doesn't matter because i know i'm getting better, i can feel i am, and also, i'm not even doing this for job opportunities since nobody in my country ( which is not china btw ), cares about cf skills. just have fun solving, and if you are not because of cheaters just take a break from cf or help the community fill their cf-cheater database. good luck <3
When I started to write problem B in Round 1109, I noticed that a gray participant had solved all the problems:) There is no doubt that he/she is a cheater:)
Let's talk about other examples. In Div2 you can see someone(often gray or green) exceeds maspy and becomes No.1, in Div1 you can see rank 2 and rank 4 are purple participantes during the contest(Round 1105), NanYan and XVIII have become legendary grandmasters, OpenAI beat all the humans in recent AWCF.
So given this, I think in your opinion all the contests aren't relevant now. But CP is still fun and challenging for human participants, so a determined lover of CP won't stop his journey, although he/she sometimes gets a negarive delta.
But as a Chinese student who's busy with research and schoolwork, I'd choose sleep early:)
Wait why did XVIII get banned but Nanyan didn't if we know to a fairly high degree of confidence both are cheaters.
You can ask the admins for help, I think I've saw the blogs that accused both of them.
In fact I'm also confused of this.
Maybe that's because XVIII is more obvious because he once participated in Div1 and solved problems from not A but G. And his ranks are higher than NanYan.
Hey man, saw your rating graph and the progress you made recently. Did you follow any specific problem set or something? I've been stuck at the same rating for quite a while. Any tips? Thank You.
I think as a specialist you can try practicing more constructive problems(they're really useful for improving your thinking ability) and practicing basic skills like easy DP, and prefix sums etc. And you can participate in other platforms' contests like Atcoder Beginner Contest or Leetcode Weekly/Biweekly.
I'm sorry if I cannot explain the methods clearly. But I have to go to sleep, as a Chinese student(Like I said above).
Just curious — what is your area of research? Asking this because I myself am into research.
I'm currently working on Machine Learning for Systems, I don't want to introduce the exact field, sorry.
If you're interested, you can search for some videos on Youtube, read some papers, and learn some known courses like CS336 from Stanford, then you may find it interesting and it's closely connected with CP in some exact topics.
I am actually into Scientific ML myself, and I myself do have a paper in NLP (arXiv:2606.16845) [no self-promotion intended btw!], that's why I am automatically impressed by anyone who say that they're doing research. I do CP as a hobby only.
But thanks for the guidance, I have not done a formal course on Language Modelling, might consider it later. I am in my final year of graduation in Mech Engg, and working on ML in Fluid Dynamics, which itself, is quite heavy-load.
Nice to meet! :)
I have built a simulator of fluid dynamics in a relevant course, that's pretty HARD for me, so I think you're so talented to research in that area.
Btw, I also do CP as a hobby only because there're a lot of participants who're more powerful than me in my school o(╥﹏╥)o .But CP is worth a lifelong hobby, I think.
Nice to meet too!:)
You literally don't need to care about who's better than you at CP XD A hobbyist being an IM is not normal at all! XD
Maybe that's because I learned MO and OI in my early years:) We tend to practice on Luogu more then.
ima keep it a buck, if someone REALLY wants to cheat with ai, if it isn't in person and not online proctored, you can't really tell if they are cheating unless its blatantly obvious. I think from here, rating still has a good amount of meaning as I placed at roughly 900 for trusted users on the last div 3 meaning that assuming all cheaters were cheating to the best of their ability, only AT MAX 10% of the people were cheating which honestly not even that bad and im assuming lower at the upper divs. Additioanlly when considering that only like a couple hundred people ak the contest from trust partiicpants and given the fact ak on this contest wasn't impossible for a div 3 partiicpant (last problem was very solvable), theres probably less than 5% of the people (by my estimates) who are cheating and thus although they may get top spots, the only negative is perhaps rating deflation but I can deal with that to some extent and if your main motivation for cf was rating in the first place, then perhaps playing a video game where there arent cheaters and you can gain ranking and doing cf purely for the fun of solving problems may be in your best interests/
Cheating mostly happens in div 3 / 4 rounds nowadays!
Why should we disregard div 3/4 rounds just because of AI cheating? Personally, I think these rounds are quite fun and educational, especially for less experienced users.
It's fun and exciting — obviously. But I think at this point, the rating is meaningless.
Auto comment: topic has been updated by SrabonGitikar (previous revision, new revision, compare).
the answer to that question is that rating will not matter in the near future for any kind of division because it's just not a trustworthy metric anymore. Sure, people like tourist don't need to prove that they are legit, but in general, you will never, ever, know if somebody got helped by ai if it's not obvious.
the better you solve tasks the higher rating you have so rating is still relevant