Live data from Hacker News

PhD Knowledge Not Required: A Reasoning Challenge for Large Language Models

arxiv.org

41–50 of 85 posts

Re: PhD Knowledge Not Required: A Reasoning Challenge for Large Language Models

#41

This doesn't feel like a "reasoning" challenge. The mental skill required to solve most of these seems to be the ability to loop over all known members of a category like "popular brand names" or "well-known actors" and see if they fit the clue. As a human, you'd expect to fail either because you didn't know a category member (e.g. as a non-American I have no idea WTF "Citgo" is; I could never get the answer to the f…

I had the same thought. It reminds me of solving Project Euler problems, where there is often an obvious naive approach which is guaranteed to produce the correct answer but would consume prohibitive memory/compute resources to execute to completion. I suspect the models would perform much better if prompted to formulate a strategy for efficiently solving these challenges rather than solving them directly… which indicates a direction for potential improvement I suppose.

Re: PhD Knowledge Not Required: A Reasoning Challenge for Large Language Models

#42

This doesn't feel like a "reasoning" challenge. The mental skill required to solve most of these seems to be the ability to loop over all known members of a category like "popular brand names" or "well-known actors" and see if they fit the clue. As a human, you'd expect to fail either because you didn't know a category member (e.g. as a non-American I have no idea WTF "Citgo" is; I could never get the answer to the f…

I agree that recall seems to play an important role in solving these problems. Similar to how the ARC-AGI problems seem to depend on visual perception of shapes and colors. When I come up with the correct answers to such puzzles, I feel subjectively that the answers flashed into my mind, not that I reasoned my way to them.

Re: PhD Knowledge Not Required: A Reasoning Challenge for Large Language Models

#44
post #37

What is so interesting to me is that the reasoning traces for these often have the correct answer, but the model fails to realize it. Problem 3 ("Dry Eye"), R1: "Wait, maybe "cubitus valgus" – no, too long. Wait, three letters each. Let me think again. Maybe "hay fever" is two words but not three letters each. Maybe "dry eye"? "Dry" and "eye" – both three letters. "Dry eye" is a condition. Do they rhyme? "Dry" (d-rye…

The other day I watched Gemini reasoning go in circles for ~30 seconds trying to determine if 9.11 was greater than 9.9. ---- "Realize a direct comparison of tenths is not enough: Wait, this is not right. In 9.9, the tenths digit is 9. In 9.11, the tenths digit is 1. This initial comparison is incorrectly stated. Let's re-examine. Re-examine tenths place comparison (correctly): In 9.9, the tenths digit is 9. In 9.11,…

I wonder if RLHF interfered with 9.11 (which could be interpreted as a date), preventing the model from considering it naturally.

Wonder if the same thing would have happened with 9.12.

What was your original prompt?

Re: PhD Knowledge Not Required: A Reasoning Challenge for Large Language Models

#45
post #30
post #24

The term "PhD Knowledge" makes me cringe. PhD is about contribution to a field, not a standardized test score or a weird status-bearing substitute for IQ.

In this case it is a shorthand for specialist level knowledge. Thre kind of information you would expect a person pursuing a PhD in a specific field to know but someone not doing that to not know. For example how in the “Humanity's Last Exam” AI evaluation dataset they have this question: “Hummingbirds within Apodiformes uniquely have a bilaterally paired oval bone, a sesamoid embedded in the caudolateral portion of…

That's a jargon and search problem at best. I'm not convinced at all.

Re: PhD Knowledge Not Required: A Reasoning Challenge for Large Language Models

#46

Are these really reasoning challenges? Seems like they are really solved via brute force or guess and check

Heuristic search, not exhaustive search, is an essential ingredient of reasoning. Has been true since chess. Remains true with MCTS, LLMs, etc.

Re: PhD Knowledge Not Required: A Reasoning Challenge for Large Language Models

#47
post #45
post #30

Earlier quoted context omitted.

In this case it is a shorthand for specialist level knowledge. Thre kind of information you would expect a person pursuing a PhD in a specific field to know but someone not doing that to not know. For example how in the “Humanity's Last Exam” AI evaluation dataset they have this question: “Hummingbirds within Apodiformes uniquely have a bilaterally paired oval bone, a sesamoid embedded in the caudolateral portion of…

That's a jargon and search problem at best. I'm not convinced at all.

> I'm not convinced at all.

What are you not convinced about?

Are you not convinced this is what they mean?

Re: PhD Knowledge Not Required: A Reasoning Challenge for Large Language Models

#48
post #37

What is so interesting to me is that the reasoning traces for these often have the correct answer, but the model fails to realize it. Problem 3 ("Dry Eye"), R1: "Wait, maybe "cubitus valgus" – no, too long. Wait, three letters each. Let me think again. Maybe "hay fever" is two words but not three letters each. Maybe "dry eye"? "Dry" and "eye" – both three letters. "Dry eye" is a condition. Do they rhyme? "Dry" (d-rye…

I hope the new models will be trained with better words to continue the thought process. Right now it seems like "wait", "but", "let me think again", etc. are the main ones, which seem to encourage self-doubt too much. They need some good balance instead.

Re: PhD Knowledge Not Required: A Reasoning Challenge for Large Language Models

#49
post #39
post #37

What is so interesting to me is that the reasoning traces for these often have the correct answer, but the model fails to realize it. Problem 3 ("Dry Eye"), R1: "Wait, maybe "cubitus valgus" – no, too long. Wait, three letters each. Let me think again. Maybe "hay fever" is two words but not three letters each. Maybe "dry eye"? "Dry" and "eye" – both three letters. "Dry eye" is a condition. Do they rhyme? "Dry" (d-rye…

It feels like lot of the reasoning tokens go to waste on pure brute force approach - plugging in numbers and evaluating and comparing against the answer. "Nope, that didn't work, let's try 4 instead of 6 this time", etc. What if the reward function instead focuses on diversity of procedures within a token budged (10k - 20k tokens). I.e. RL rewards the model in trying different methods or generating different hypothes…

I would say that diversity isn't something that's easy to reenforce, but I do think it will occur as a natural consequence of optimizing for shorter chains of thought according to a wide variety of problems. Of course, the nature of the data may lead it to do brute force, but that can be fixed with clever fine tuning.

Re: PhD Knowledge Not Required: A Reasoning Challenge for Large Language Models

#50
post #37

What is so interesting to me is that the reasoning traces for these often have the correct answer, but the model fails to realize it. Problem 3 ("Dry Eye"), R1: "Wait, maybe "cubitus valgus" – no, too long. Wait, three letters each. Let me think again. Maybe "hay fever" is two words but not three letters each. Maybe "dry eye"? "Dry" and "eye" – both three letters. "Dry eye" is a condition. Do they rhyme? "Dry" (d-rye…

The other day I watched Gemini reasoning go in circles for ~30 seconds trying to determine if 9.11 was greater than 9.9. ---- "Realize a direct comparison of tenths is not enough: Wait, this is not right. In 9.9, the tenths digit is 9. In 9.11, the tenths digit is 1. This initial comparison is incorrectly stated. Let's re-examine. Re-examine tenths place comparison (correctly): In 9.9, the tenths digit is 9. In 9.11,…

I wonder how much of this arithmetic confusion comes from software versioning numbers where the decimal character is actually a tuple-delimiter.
Post reply on HN