OpenAI IMO Gold - /sci/ (#16728408) [Archived: 62 hours ago]

Anonymous
7/19/2025, 3:26:10 PM No.16728408
Screenshot From 2025-07-19 15-23-03
Screenshot From 2025-07-19 15-23-03
md5: 1bffea206ffff1c6c9cb51ff5e428218๐Ÿ”
It's over.

Do you find this intimidating or encouraging?
Replies: >>16728526 >>16728531 >>16728535 >>16728830 >>16729295
Anonymous
7/19/2025, 6:02:38 PM No.16728526
>>16728408 (OP)
>computer can do math
a surprise to be sure
Anonymous
7/19/2025, 6:02:49 PM No.16728527
SAAR it is not redeemable for many a months. Saar please understand. Redeem the gpt5 very good still Saar
Anonymous
7/19/2025, 6:03:16 PM No.16728530
Good, can we now stop doing those stupid math competitions?
Anonymous
7/19/2025, 6:04:35 PM No.16728531
>>16728408 (OP)
Suddenly, pajeets will start doing well in IMO. Like they have been cheating in chess and programming interviews.
Replies: >>16728823
Anonymous
7/19/2025, 6:07:18 PM No.16728535
IMG_6734
IMG_6734
md5: c418fb9ebe14482477f68df9d06f4db1๐Ÿ”
>>16728408 (OP)
imagine being a fucking math geek right now
Anonymous
7/19/2025, 10:58:33 PM No.16728823
>>16728531
Pajeets literally invented chess.
Anonymous
7/19/2025, 11:05:25 PM No.16728830
>>16728408 (OP)
apparently they announced it at 1 am on a saturday bc deepmind got one too lol
Anonymous
7/20/2025, 2:39:03 PM No.16729291
garymarcus
garymarcus
md5: eca0bb7b6929ef78980e6f9dd2ddd395๐Ÿ”
hoomans obsolete
Anonymous
7/20/2025, 2:43:55 PM No.16729295
>>16728408 (OP)
I don't believe it. Earlier this year OpenAI got about 5% on the 2025 USA math Olympiad, now 6 months later they have a gold medal? Nah son, this OpenAI poisoning their training data again, like they do with code forces to declare that they're in the top 0.0001% of competitors.

https://arxiv.org/abs/2503.21934
Replies: >>16729329 >>16729424
Anonymous
7/20/2025, 3:45:16 PM No.16729329
>>16729295
It doesn't matter. They are just overfit. Like using a test dump to answer questions.
Anonymous
7/20/2025, 4:56:26 PM No.16729424
>>16729295
the reasoning steps seem atypical, some speculate the used a diffusion llm.
whatever it is, it doesnโ€™t use a harness like deepmind.