Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

30% drop in O1-preview accuracy when Putnam problems are slightly variated

openreview.net

Post reply on HN
Print view View original

461–470 of 558 posts

Re: 30% drop in O1-preview accuracy when Putnam problems are slightly variated

#461

Post by YetAnotherNick »

boroboro4 wrote: post #422
YetAnotherNick wrote: post #195

Earlier quoted context omitted.

I don't know why I and you are getting downvoted. Sometimes, HN crowd is just unhinged against AI.

These models are trained in two steps: training base model and then uptraining it. First step includes as much data as possible, everything company can find. For Llama models it's 15T tokens, which is ~40 TB of data. No-one really puts an effort on splitting this data into train/test/eval (and it's not very achievable either). It's just as much data as possible. So it's like 99.9999999% wrong to assume something publ…

There are benchmarks that are decided beforehand and similar sentences are removed from even the first stage of training. This is useful for tracking model performance and comparing different choices. e.g. see section 'Contamination of downstream tasks' of [1].

Every decent AI lab does this, else the benchmark result couldn't be trusted. OpenAI publishes results of ~20 benchmarks[2] and it is safe to assume they have made reasonable attempt to remove it from training set

[1]: https://arxiv.org/pdf/2107.06499

[2]: https://openai.com/index/hello-gpt-4o/

Re: 30% drop in O1-preview accuracy when Putnam problems are slightly variated

#462

Post by kevhito »

jorvi wrote: post #332
NewsaHackO wrote: post #211

Earlier quoted context omitted.

This happens literally every time. Someone always says "ChatGPT can't do this!", but then when someone actually runs the example, chatGPT gets it right. Now what the OP is going to do next is proceed to move goalposts and say like "but umm I just asked chatgpt this, so clearly they modified the code in realtime to get the answer right"

Prompt: In the Netherlands, in terms of drinks, is there a particular spirit that represents the country? > Yes, in the Netherlands, jenever (also known as genever) is the traditional spirit that represents the country. Jenever is a type of Dutch gin that has a distinctive flavor, often made from malt wine and flavored with juniper berries. It has a long history in the Netherlands, dating back to the 16th century, an…

"This is completely wrong." Is it completely wrong? Or maybe "somewhat" wrong? Maybe just lacking nuance? I know nothing about the answer to this question, so this is an honest question.

Using just a plain old search engine, for things like "national drink of the netherlands" and simlar queries, I am directed to Wikipedia's Jenever page as the top hit, and Wikipedia's list of national drinks lists Jenever and Heineken as the entries for the Netherlands. Search engines also give page after page of travel guides and blog posts, most of which list Jenever at or near the top of of their listings. One travel guide calls it "the most famous Dutch spirit and most famous Amsterdam liquor, Jenever, also spelled Genever or simply Dutch gin."

Re: 30% drop in O1-preview accuracy when Putnam problems are slightly variated

#463

Post by boroboro4 »

whimsicalism wrote: post #439
boroboro4 wrote: post #436

Earlier quoted context omitted.

What test set is being talked about here? Why does it matter what’s on this set?

the point is that putnam was never a test/benchmark being used by OpenAI or anyone else, so there is no smoking gun if you find putnam on the train set nor is it cheating or nefarious because nobody ever claimed otherwise. this whole notion of putnam as test being trained on is a fully invented grievance read the entire thread in this context

I've read the thread and I think it's not very coherent overall, I also not sure if we disagree =)

I agree that having putnam problems on OpenAI training set is not a smoking gun, however it's (almost) certain they are on training set, and having them would affect performance of the model on them too. Hence research like this is important, since it shows that observed behavior of the models is memoization to large extent, and not necessarily generalization we would like it to be.

Re: 30% drop in O1-preview accuracy when Putnam problems are slightly variated

#464

Post by gus_massa »

tame3902 wrote: post #199
gus_massa wrote: post #182

Earlier quoted context omitted.

I didn't downvote it, but short comments are a very big risk. People may misinterpret it, or think it's crackpot theory or a joke and then downvote. When in doubt, add more info, like: But the complete equation is E=sqrt(m^2c^4+p^2) that is reduced to E=mc^2 when the momentum p is 0. More info in https://en.wikipedia.org/wiki/Mass%E2%80%93energy_equivalenc...

What I learnt is that there is a rest mass and a relativistic mass. The m in your formula is the rest mass. But when you use the relativistic mass E=mc² still holds. And for the rest mass I always used m_0 to make clear what it is.

I should have used m_0 to avoid confussion. Anyway, as he sibling comment says, most modern advanced books of special relativity try to avoid the relativistic mass. It's useful for some calculations, like synchrotron, but the problem is that for forward/backward acceleration you must use other number so the relativistic mass add confussion. https://en.wikipedia.org/wiki/Mass_in_special_relativity#His...

Re: 30% drop in O1-preview accuracy when Putnam problems are slightly variated

#465

Post by ploxiln »

HenriTEL wrote: post #411
themgt wrote: post #375

Earlier quoted context omitted.

I just asked Claude 3.5 Sonnet, which appears to have improved its response with CoT but there's mistakes that demonstrate the model doesn't really "understand": Q: A woman and her son are in a car accident. The woman is sadly killed. The boy is rushed to hospital. When the doctor sees the boy he says "I can't operate on this child, he is my son". How is this possible? C: Let me think about this step by step: A woman…

I must be to tired as I can't find any flaw in that reasoning.

> The doctor is the boy's father. This is a classic riddle that highlights gender bias - many people assume doctors must be men, but don't initially consider that the doctor could be the father.

It identified the pattern, but it didn't understand the father is a man while the mother is a woman. If people assume doctors must be men, they wouldn't consider that the doctor could be the mother.

(Side note: this is an antiquated riddle/joke ... for me in particular because my mother is a doctor, but in general around half of new doctors are female, and think about Grey's Anatomy and other shows.)

Re: 30% drop in O1-preview accuracy when Putnam problems are slightly variated

#466

Post by sudosysgen »

whimsicalism wrote: post #171
fldskfjdslkfj wrote: post #164

Earlier quoted context omitted.

I dont think codeforce claims to contain novel unpublished problems. But i'm not saying it's what they did, just that it's a possibility that should be considered till/if it is debunked.

frankly i’m not sure what standard you would possibly consider a debunking codeforces constantly adds new problems that’s like the entire point of the contest, no?

OpenAI never actually directly participated in the contests. OpenAI says they took (unspecified) problems and "simulated" the score it would have had.

If they solved recent contests in a realistic contest simulation I would expect them to give the actual solutions and success rates as well, like they did for IOI problems, so I'm actually confused as to why they didn't.

Re: 30% drop in O1-preview accuracy when Putnam problems are slightly variated

#467

Post by CamperBob2 »

whimsicalism wrote: post #443
CamperBob2 wrote: post #434

Earlier quoted context omitted.

o1-pro just answered (it often takes a few minutes). Not sure if this is any better (is it?) but at least it is starting to admit when it isn't sure about something, rather than just spewing BS confidently. That seems like a genuine win. ---------snip-------- From the handwriting and context, this appears to be a mid-19th-century petition submitted to the Orphans’ Court of Baltimore County. Below is a point‐by‐point…

o1-pro takes images? had no idea

Yes, it will take .JPGs but not .PDFs. I thought it didn't support either until I checked just now.

Re: 30% drop in O1-preview accuracy when Putnam problems are slightly variated

#468

Post by Clubber »

roughly wrote: post #369
wongarsu wrote: post #361

Earlier quoted context omitted.

We are pretty certain that humans can reason, yet they are sometimes wrong. Even if you give them the same problem over and over again with slight variations. LLMs get things wrong due to different factors than humans (humans lose focus, LLMs have randomness applied when sampling their responses to improve results). But clearly we have to choose a goal somewhat below 100% if we want a test that doesn't conclude that…

The difference is we _know_ that LLMs are fancy stochastic models, we don't know that they're capable of reasoning, and the null hypothesis is that they're not (because we know what they _are_ - we built them) - any "reasoning" is an emergent property of the system, not something we built them to do. In that case, evidence they're not reasoning - evidence they're stochastic parrots doing a performance of reasoning -…

>we don't know that they're capable of reasoning

Apple AI researchers released a paper on it. They say no.

https://arxiv.org/pdf/2410.05229

Re: 30% drop in O1-preview accuracy when Putnam problems are slightly variated

#469

Post by cscurmudgeon »

falcor84 wrote: post #451
fn-mote wrote: post #102

Earlier quoted context omitted.

The point is that the “good student” will still do well on the variations, not suffer a 30% decrease in grade.

I'm not following, why would you assume that a good student taking an exam at the edge of their ability not do significantly better if they trained on the exact same questions (with the same solutions), as opposed to ones that are slightly different? I for one have absolutely struggled as a student when faced with questions that seemed similar to ones from previous exams, but actually had a crucial difference, and by…

Because good students don’t cram. It is not optimal. Understanding the underlying structure is much easier than surface-level cramming.

Re: 30% drop in O1-preview accuracy when Putnam problems are slightly variated

#470

Post by whimsicalism »

boroboro4 wrote: post #463
whimsicalism wrote: post #439

Earlier quoted context omitted.

the point is that putnam was never a test/benchmark being used by OpenAI or anyone else, so there is no smoking gun if you find putnam on the train set nor is it cheating or nefarious because nobody ever claimed otherwise. this whole notion of putnam as test being trained on is a fully invented grievance read the entire thread in this context

I've read the thread and I think it's not very coherent overall, I also not sure if we disagree =) I agree that having putnam problems on OpenAI training set is not a smoking gun, however it's (almost) certain they are on training set, and having them would affect performance of the model on them too. Hence research like this is important, since it shows that observed behavior of the models is memoization to large ex…

nobody serious (like OAI) was using the putnam problems to claim generalization. this is a refutation in search of a claim - and many people in the upstream thread are suggesting that OAI is doing something wrong by training on a benchmark.

OAI uses datasets like frontiermath or arc-agi that are actually held out to evaluate generalization.

Post reply on HN