Note: The text editorials will be provided by the authors of the round. This video tutorial acts as an additional resource for those who prefer video over text, not as a substitute for the text editorial.
During today’s contest solution discussion, we tested ChatGPT o3-mini-high on our contest problems. And get this: it solved problems C and D! It claimed that it did it without internet access.
We asked about its capability, and it said that if it made a Codeforces account, its rating would be 3500–4000! I don’t know if AI has learned to solve very hard problems or if it’s getting overconfident.
All I can say is, things are getting pretty SCARY. I have no idea what’s going to happen with CP contests and competitive programmers soon.
2059A — Milya and Two Arrays
2059B — Cost of the Array
2059C — Customer Service
GPT o3-mini-high Solved C
2059D — Graph and Graph
GPT o3-mini-high Solved D
Asking ChatGPT if it is Cheating (Has Internet Access)
o3 claiming that it has rating 3500 on Codeforces!
Trying to Solve E2 with o2 — Not Successful
P.S. Seems like rank 4000+ should be called GPT soon.
Since the AI models are getting better at problem solving , why the plag checker isn't getting better? if the plag checkers become better it would solve most problems of cheating with ai or cheating in general. also maybe making rules about cheating a bit more strict would be a good step
this is a highly misleading blog post title