[0x20cowboy]

> LLMs are PhD-level reasoners in math and science, yet they fail at children's puzzles. How is this possible?

Because they are not.

Pattern matching questions on a contrived test is not the same thing as understanding or reasoning.

It’s the same reason why most of the people who pass your leetcode tests don’t actually know how to build anything real. They are taught to the test not taught to reality.

[gwd]

> Pattern matching questions on a contrived test is not the same thing as understanding or reasoning.

Do submarines swim? I don't really care if it gets me where I want to go. The fact is that just two days ago, I asked Claude to look at some reasonably complicated concurrent code to which I had added a new feature, and asked it to list what tests needed to be added; and then when I asked GPT-5 to add them, it one-shot nailed the implementations. I've written a gist of it here:

https://gitlab.com/-/snippets/4889253

Seriously just even read the description of the test it's trying to write.

In order to one-shot that code, it had to understand:

- How the cache was supposed to work

- How conceptually to set up the scenario described

- How to assemble golang's concurrency primitives (channels, goroutines, and waitgroups), in the correct order, to achieve the goal.

Did it have a library of concurrency testing patterns in its head? Probably -- so do I. Had it ever seen my exact package before in its training? Never.

I just don't see how you can argue with a straight face that this is "pattern matching". If that's pattern matching, then pattern matching is not an insult.

If anything, the examples in this article are the opposite. Take the second example, which is basically 'assemble these assorted pieces into a rectangle'. Nearly every adult has assembled a minimum of dozens of things in their lives; many have assembled thousands of things. So it's humans in this case who are simply "pattern matching questions on a contrived test", and the LLMs, which almost certainly didn't have a lot of "assemble these items" in their training data, that are reasoning out what's going on from first principles.

[HarHarVeryFunny]

> Do submarines swim?

It doesn't matter HOW LLMs "swim" as long as they can, but the point being raised is whether they actually can.

It's as if LLMs can swim in the ocean, in rough surf, but fail to swim in rivers or swimming pools, because they don't have a generalized ability to swim - they've just been RL-trained on the solution steps to swimming in surf, but since those exact conditions don't exist in a river (which might seem like a less challenging environment), they fail there.

So, the question that might be asked is when LLMs are trained to perform well in these vertical domains like math and programming, where it's easy to verify results and provide outcome- or process-based RL rewards, are they really learning to reason, or are they just learning to pattern match to steer generation in the direction of problem-specific reasoning steps that they had been trained on?

Does the LLM have the capability to reason/swim, or is it really just an expert system that has been given the rules to reason/swim in certain cases, but would need to be similarly hand fed the reasoning steps to be successful in other cases?

I think the answer is pretty obvious given that LLM's can't learn at runtime - can't try out some reasoning generalization they may have arrived at, find that it doesn't work in a specific case, then explore the problem and figure it out for next time.

Given that it's Demis Hassabis who it pointing out this deficiency of LLMs (and has a 5-10 year plan/timeline to fix it - AGI), not some ill-informed LLM critic, it seems silly to deny it.

[schrectacular]

>> Do submarines swim?

>It doesn't matter HOW LLMs "swim" as long as they can, but the point being raised is whether they actually can.

>It's as if LLMs can swim in the ocean, in rough surf, but fail to swim in rivers or swimming pools

Just like submarines!

[MobiusHorizons]

What? Submarines can definitely “swim” in rivers, although shallow water is certainly more challenging for a submerged vessel. Most submarines are a bit big for most swimming pools, but small ones like ROVs are frequently tested in pools.

[gwd]

> I think the answer is pretty obvious given that LLM's can't learn at runtime - can't try out some reasoning generalization they may have arrived at, find that it doesn't work in a specific case, then explore the problem and figure it out for next time.

This is just a problem of memory. Supposing that an LLM did generate a genuinely novel insight, it could in theory they could write a note for itself so that next time they come online, they can read through a summary of the things they learned. And it could also write synthetic training data for itself so that the next time they're trained, that gets incorporated into its general knowledge.

OpenAI allows you to fine-tune GPT models, I believe. You could imagine a GPT system working for 8 hours in a day, then spending a bunch of time looking over all its conversation looking for patterns or insights or things to learn, and then modifying its own fine-tuning data (adding, removing, or modifying as appropriate), which it then used to train itself overnight, waking up the next morning having synthesized the previous day's experience.

[HarHarVeryFunny]

> This is just a problem of memory

How does memory (maybe later incorporated via fine tuning) help if you can't figure out how to do something in the first place ?

That would be a way to incorporate new declarative data at "runtime" - feedback to the AI intern as to what it is doing wrong. However, in order to do something effectively by yourself generally requires more than just new knowledge - it requires personal practice/experimentation etc, since you need to learn how to act based on the contents of your own mind, not that of the instructor.

Even when you've had enough practice to become proficient at a taught skill, you may not be able to verbalize exactly what you are doing (which is part of the teacher-student gap), so attempting to describe then capture that as textual/context "sensory input" is not always going to work.

[naasking]

> are they really learning to reason, or are they just learning to pattern match to steer generation in the direction of problem-specific reasoning steps that they had been trained on?

Are you sure there's a real difference? Do you have a definition of "reasoning" that excludes this?

[gwd]

So I do think there are two distinct types of activities involved in knowledge work:

1. Taking established techniques or concepts and appropriately applying them to novel situations.

2. Inventing or synthesizing new, never-before-seen techniques or concepts

The vast majority of the time, humans do #1. LLMs certainly do this in some contexts as well, as demonstrated by my example above. This to me counts as "understanding" and "thinking". Some people define "understanding" such that it's something only humans can do; to which I respond, I don't care what you call it, it's useful.

Can LLMs do #2? I don't know. They've got such extensive experience that how would you know if they'd invented a technique vs had seen it somewhere?

But I'd venture to argue that most humans never or rarely do #2.

[HarHarVeryFunny]

> But I'd venture to argue that most humans never or rarely do #2.

That seems fair, although the distinction between synthesizing something new and combining existing techniques is a bit blurry.

What's missing from LLMs though is really part of 1). If techniques A, B, C & D are all the tools you need to solve a novel problem, then a human has the capability of learning WHEN to use each of these tools, and in what order/combination, to solve that problem - a process of trial and error, generalization and exception, etc. It's not just the techniques (bag of tools) you need, but also the rules (acquired knowledge) of how they can be used to solve different problems.

LLMs aren't able to learn at runtime from their own experience, so the only way they can learn these rules of when to apply given tools (aka reasoning steps) - is by RL training on how they have been successfully used to solve a range of problems in the training data. So, the LLM may have learnt that in specific context it should first apply tool A (generate that reasoning step), etc, etc, but that doesn't help it to solve a novel problem where the same solution step selection process doesn't apply, even if the tools A-D are all it needs (if only it could learn how to apply them to this novel problem).

[mjr00]

It's trivial to demonstrate that LLMs are pattern matching rather than reasoning. A good way is to provide modified riddles-that-aren't. As an example:

> Prompt: A man working at some white collar job gets an interview scheduled with an MBA candidate. The man says "I can't interview this candidate, he's my son." How is this possible?

> ChatGPT: Because the interviewer is the candidate’s mother. (The riddle plays on the assumption that the interviewer must be a man.)

This is clearly pattern matching and overfitting to the "doctor riddle" and a good demonstration of how there's no actual reasoning going on. A human would read the prompt and initially demonstrate confusion, which LLMs don't demonstrate because they don't actually reason.

[Workaccount2]

Over fitting isn't evidence of non-reasoning, but that aside, what's interesting is that ChatGPT (free) trips on this, as did older models. But GPT-5 thinking, Opus 4, and Gemini 2.5 Pro all pointed out that there is no trick and it's likely the man just views it as a conflict of interest to interview his son.

It's hard to say whether this has been trained out (it's an old example) or if it's just another hurdle that general model progression has overcome.

[2ap]

OK. But, in Claude Sonnet 4:

'This is possible because the man is the candidate's father. When he says "he's my son," he's simply stating their family relationship. The scenario doesn't present any logical contradiction - a father could very well be in a position where he's supposed to interview his own son for a job. This would create a conflict of interest, which is why he's saying he can't conduct the interview. It would be inappropriate and unfair for a parent to interview their own child for a position, so he would need to recuse himself and have someone else handle the interview. The phrasing might initially seem like it's setting up a riddle, but it's actually a straightforward situation about professional ethics and avoiding conflicts of interest in hiring.'

EDIT - this is described better by other posters.

[naasking]

> It's trivial to demonstrate that LLMs are pattern matching rather than reasoning.

Again, this is just asserting the premise that reasoning cannot include pattern matching, but this has never been justified. What is your definition for "reasoning"?

> This is clearly pattern matching and overfitting to the "doctor riddle" and a good demonstration of how there's no actual reasoning going on.

Not really, no. "Bad reasoning" does not entail "no reasoning". Your conclusion is simply too strong for the evidence available, which is why I'm asking for a rigourous definition of reasoning that doesn't leave room for disagreement about whether pattern matching counts.

[mjr00]

If your assertion is that you can't prove reasoning isn't just pattern matching, then I counter by saying you can't prove reasoning isn't just chaining a large number of IF/THEN/ELSE logic statements and therefore computers have been generally intelligent since ~1960.

[DenisM]

We kinda move from the situation “LLM can only do what it seen before” to “LLM can do something by composing several things it has seen before”. We didn’t get to the situation “LLM can do things it has not seen before”.

The practicality of the situation is that a lot of problems fall into the second bucket. We all like to think we deal with novel problems, but most of what we can think of was already considered by another human and captured by llm. You had to invent something deliberately unique, and that’s telling. Most startup ideas are invented more than once, for example.

The key shortcoming of the llm is that it is not aware of its own limits. If it ever becomes aware it can outsource such rare things to mechanical Turk.

[adastra22]

I routinely use LLMs to do things that have never been done before. It requires carefully structured prompting and context management, but it is quite doable.

[adastra22]

People make the same sort of mistakes.

[mjr00]

Please explain how this is relevant to the topic at hand. Thanks!

[HarHarVeryFunny]

I define intelligence as prediction (degree of ability to use past experience to correctly predict future action outcomes), and reasoning/planning as multi-step what-if prediction.

Certainly if a human (or some AI) has learned to predict/reason over some domain, then what they will be doing is pattern matching to determine the generalizations and exceptions that apply in a given context (including a hypothetical context in a what-if reasoning chain), in order to be able to select a next step that worked before.

However, I think what we're really talking about here isn't the mechanics of applying learnt reasoning (context pattern matching), but rather the ability to reason in the general case, which requires the ability to LEARN to solve novel problems, which is what is missing from LLMs.

A system that has a fixed set of (reasoning/prediction) rules, but can't learn new ones for itself, seems better regarded as an expert system. We need to make the distinction between a system that can only apply rules, and one that can actually figure out the rules in the first place.

In terms of my definitions of intelligence and reasoning, based around ability to use past experience to learn to predict, then any system that can't learn from fresh experience doesn't meet that definition.

Of course in humans and other intelligent animals the distinction between past and ongoing experience doesn't apply since they can learn continually and incrementally (something that is lacking from LLMs), so for AI we need to use a different vocabulary, and "expert system" seems the obvious label for something that can use rules, but not discover them for itself.

[naasking]

> but rather the ability to reason in the general case, which requires the ability to LEARN to solve novel problems, which is what is missing from LLMs.

I don't think it's missing, zero shot prompting is quite successful in many cases. Maybe you find the extent that LLMs can do this to be too limited, but I'm not sure that means they don't reason at all.

> A system that has a fixed set of (reasoning/prediction) rules, but can't learn new ones for itself, seems better regarded as an expert system.

I think expert systems are a lot more limited than LLMs, so I don't agree with that classification. LLMs can generate output that's out of distribution, for instance, which is not something that's classic expert systems can do (even if you think LLM OOD is still limited compared to humans).

I've elaborated in another comment [1] what I think part of the real issue is, and why people keep getting tripped up by saying that pattern matching is not reasoning. I think it's perfectly fine to say that pattern matching is reasoning, but pattern matching has levels of expressive power. First-order pattern matching is limited (and so reasoning is limited), and clearly humans are capable of higher order pattern matching which is Turing complete. Transformers are also Turing complete, and neural networks can learn any function, so it's not a matter of expressive power, in principle.

Aside from issues stemming from tokenization, I think many of these LLM failures are because they aren't trained in higher order pattern matching. Thinking models and the generalization seen from grokking are the first steps on this path, but it's not quite there yet.

[1] https://news.ycombinator.com/item?id=45277098

[HarHarVeryFunny]

Powerful pattern matching is still just pattern matching.

How is an LLM going to solve a novel problem with just pattern matching?

Novel means it has never seen it before, maybe doesn't even have the knowledge needed to solve it, so it's not going to be matching any pattern, and even if it did, that would not help if it required a solution different to whatever the pattern match had come from.

Human level reasoning includes ability to learn, so that people can solve novel problems, overcome failures by trial and error, exploration, etc.

So, whatever you are calling "reasoning" isn't human level reasoning, and it's therefore not even clear what you are trying to say? Maybe just that you feel LLMs have room for improvement by better pattern matching?

[freejazz]

It seems readily apparent there is a difference given their inability to do tasks we would otherwise reasonably describe as achievable via basic reasoning on the same facts.

[naasking]

I agree LLMs have many differences in abilities relative to humans. I'm not sure what this implies for their ability to reason though. I'm not even sure what examples about their bad reasoning can prove about the presence or absence of any kind of "reasoning", which is why I keep asking for definitions to remove the ambiguity. If examples of bad reasoning sufficed, then this would prove that humans can't reason either, which is silly.

A rigourous definition of "reasoning" is challenging though, which is why people consistently can't provide a general one that's satisfactory when I ask, and this is why I'm skeptical that pattern matching isn't a big part of it. Arguments that LLMs are "just pattern matching" are thus not persuasive arguments that they are not "reasoning" at some cruder level.

Maybe humans are just higher order pattern matchers and LLMs are only first or second-order pattern matchers. Maybe first-order pattern matching shouldn't count as "reasoning", but should second-order? Third-order? Is there evidence or some proof that LLMs couldn't be trained to be higher order pattern matchers, even in principle?

None of the arguments or evidence I've seen about LLMs and reasoning is rigourous or persuasive on these questions.

[freejazz]

Nothing about the uncertainty of the definition for 'reasoning' requires that pattern matching be part of the definition.

[Grimblewald]

> just read the description

Seems fine enough to me. Wanna really challenge an LLM? get it to make an image stitching algorithm that isn't shit. Implement the results from brown et al https://link.springer.com/article/10.1007/s11263-006-0002-3 and I'll be impressed.

This is a paper from 2007 and there are plenty of packages available to help make it all happen through some API calls and a bit of cleverness on the coders part, and so far not a single LLM has gotten close to an acceptable implementation. Not a single one.

Now, why is it so hard? Because there's not public code for good quality high performance image stitching on the level of the image composite editor microsoft research once hosted. There's nothing for the LLM's to draw on and they fundamentally lack reasoning / planning other than something that superficially resembles it, but it falls apart for out of domain things where humans still do fine even if new to the task.

[washadjeffmad]

I tend to skip these threads because there's a gulf in understanding between those who have accepted the marketing and those who are hoping to profit from them.

"PhD level reasoning" means they have asked PhDs questions and finetuned on their responses. It does not mean that every response is "PhD level", that the models only provide responses that PhDs have validated, or that every response is correct. It is "Lysol kills 99.9% of germs" logic.

The domains I work in and contribute to largely view being paid to write out their thought processes when answering questions models answer incorrectly as a novelty, like grading homework with some minor bragging rights.

[OtherShrezzing]

>and the LLMs, which almost certainly didn't have a lot of "assemble these items" in their training data

I don't think this assumption is sound. Humans write a huge amount on "assemble components x and y to make entity z". I'd expect all LLMs to have consumed every IKEA type instruction manual, the rules for Jenga, all geometry textbooks and papers ever written.

[amelius]

Most of our coding is just plumbing. Getting data from one place to where it needs to be. There is no advanced reasoning necessary. Just a good idea of the structure of the code and the data-structures.

Even high school maths tests are way harder than what most professional programmers do on a daily basis.

[vlovich123]

I could be mistaken but generally LLMs cannot tackle out-of-domain problems whereas humans do seem to have that capability. Relatedly, the energy costs are wildly different suggesting that LLMs are imitating some kind of thought but not simulating it. They’re doing a remarkable job of passing the Turing test but that says more about the limitations of the Turing test than it does about the capabilities of the LLMs.

[Akronymus]

> I just don't see how you can argue with a straight face that this is "pattern matching". If that's pattern matching, then pattern matching is not an insult.

IMO its still "just" a, very good, autocomplete. No actual reasoning, but lots of statistics on what is the next token to spit out.

[NoahZuniga]

> Do submarines swim?

That's the main point of the parent comment. Arguing about the definition of "reasoning" or "pattern matching" is just a waste of time. What really matters is if it produces helpful output. Arguing about that is way better!

Instead of saying: "It's just pattern matching -> It won't improve the world", make an argument like: "AI's seem to have trouble specializing like humans -> adopting AI will increase error rates in business processes -> due to the amount of possible edge cases, most people will get into an edge case with no hope of escaping it -> many people's lives will get worse".

The first example relies on us agreeing on the definition of pattern matching, and then taking a conclusion based on how those words feel. This has no hope of convincing me if I don't like your definition! The second one is an argument that could potentially convince me, even if I'm an AI optimist. It is also just by itself an interesting line of reasoning.

[ozgung]

No it's not "just a very good autocomplete". I don't know why people repeat this thing (it's wrong) but I find it an extremely counterproductive position. Some people just love to dismiss the capabilities of AI with a very shallow understanding of how it works. Why?

It generates words one by one, like we all do. This doesn't mean it does just that and nothing else. It's the mechanics of how they are trained and how they do inference. And most importantly how they communicate with us. It doesn't define what they are or their limits. This is reductionism. Ignoring the mathematical complexity of a giant neural network.

[Bjartr]

> like we all do

Do we though? Sure, we communicate sequentially, but that doesn't mean that our internal effort is piecewise and linear. A modern transformer LLM however is. Each token is sampled from a population exclusively dependent on the tokens that came before it.

Mechanistically speaking, it works similarly to autocomplete, but at a very different scale.

Now how much of an unavoidable handicap this incurs, if any, is absolutely up for debate.

But yes, taking this mechanistic truth and only considering it in a shallow manner underestimates the capability of LLMs by a large degree.

[kenjackson]

Our thinking is also based only on events that occurred previously in time. We don’t use events in the future.

[ElevenLathe]

Is this a certainty? I thought it was an open question whether quantum effects are at play in the brain, and those have a counterintuitive relationship with time (to vastly dumb things down in a way my grug mind can comprehend).

[freejazz]

This is unhelpfully obtuse

[CamperBob2]

I don't know why people repeat this thing (it's wrong)

Because they simply don't care if they're wrong. At this point, given what we've seen, that seems like the only explanation left.

You don't need to be a fanatical AGI evangelist, but when an "autocomplete" starts winning international math competitions, you need to start calling it something else.

[karmakaze]

I can't say for certain that our wetware isn't "just a very good autocomplete".

[esafak]

A very good autocomplete is realized by developing an understanding.

[faangguyindia]

>Pattern matching questions on a contrived test is not the same thing as understanding or reasoning.

I think most of the problem i solve is also a pattern matching. The problems i am good at solving are the ones i've seen before or the ones i can break into problems i've seen before.

[artrockalter]

> It’s the same reason why most of the people who pass your leetcode tests don’t actually know how to build anything real. They are taught to the test not taught to reality.

True, and "Agentic Workflows" are now playing the same role as "Agile" in that both take the idea that if you have many people/LLMs that can solve toy problems but not real ones then you can still succeed by breaking down the real problems into toy problems and assigning them out.

[Grimblewald]

I fully agree, and any doubters, I can assure you that an LLM is bested by most undergraduates when it comes to reasoning. LLM's will universally get smoked by any PhD. They have great big wealth of knowledge to draw on but critical thinking is sorely lacking.

LLM's strength is being an interactive encyclopedia, not a decision making thing.

[ACCount37]

"Not understanding or reasoning" is anthropocentric cope. There is very little practical difference between "understanding" and "reasoning" implemented in human mind and that implemented in LLMs.

One notable difference, however, is that LLMs disproportionately suck at spatial reasoning. Which shouldn't be surprising, considering that their training datasets are almost entirely text. The ultimate wordcel makes for a poor shape rotator.

All ARC-AGI tasks are "spatial reasoning" tasks. They aren't in any way special. They just force LLMs to perform in an area they're spectacularly weak at. And LLMs aren't good enough yet to be able to brute force through this innate deficiency with raw intelligence.

[HighGoldstein]

> There is very little practical difference between "understanding" and "reasoning" implemented in human mind and that implemented in LLMs.

Source?

[ACCount37]

The primary source is: measured LLM performance on once-human-exclusive tasks - such as high end natural language processing or commonsense reasoning.

Those things were once thought to require a human mind - clearly, not anymore. Human commonsense knowledge can be both captured and applied by a learning algorithm trained on nothing but a boatload of text.

But another important source is: loads and loads of mech interpret research that tried to actually pry the black box open and see what happens on the inside.

This found some amusing artifacts - such as latent world models that can be extracted from the hidden state, or neural circuits corresponding to high level abstracts being chained together to obtain the final outputs. Very similar to human "abstract thinking" in function - despite being implemented on a substrate of floating point math and not wet meat.

[freejazz]

I haven't seen LLMs perform common sense reasoning. Feel free to share some links. Your post reads like anthropomorphized nonsense.

[keeda]

One of the most astonishing things about LLMs is that they actually seem to have achieved general common-sense reasoning to a signficant extent. Example from the thread about somebody ordering 18000 waters at a drive-through: https://news.ycombinator.com/item?id=45067653

TL;DR: Even without being explicitly prompted to, a pretty weak LLM "realized" that a thousand glasses of water was an unreasonable order. I'd say that's good enough to call "common sense".

You can try it out yourself! Just pick any AI chatbot, make up situations with varying levels of absurdity, maybe in a roleplay setting (e.g. "You are a fast food restaurant cashier. I am a customer. My order is..."), and test how it responds.

[ACCount37]

What? Do you even know what "commonsense reasoning" means?

[freejazz]

Do you?

[Workaccount2]

There is no source and arguing this is dumb because no one knows what reasoning or understanding is. No one.

So all we have is "Does it swim like a duck, look like a duck, quack like a duck?"

[adastra22]

I’m sympathetic to your point, but this isn’t quite fair. The field of psychology does exist.

[Workaccount2]

Neuroscience is the field that would be closest to this. But even they are empty handed with evidence and heavy with hypotheses.

[adastra22]

No, psychology is right. Psychology studies what the properties of thought are. Neuroscience studies the specific biochemical mechanisms of the brain. Psychology is the study of what mental reasoning IS, while neuroscience is the study of HOW neurons in our brain implement it.

If you are asking “ok, but what is reasoning, really? What definition of reasoning would enable us to recognize whether it is going on in this AI or not?” it is a question of psychology. Unless we are restricting ourselves to whole brain emulation only.

[NooneAtAll3]

...literally benchmarks the post is all about?

practical difference is about results - and results are here

[dwallin]

Very much agree with this. Looking at the dimensionality of a given problem space is a very helpful heuristic when analyzing how likely an llm is going to be suitable/reliable for that task. Consider how important positional encodings are LLM performance. You also then have an attention model that operates in that 1-dimensional space. With multidimensional data significant transformations to encode into a higher dimensional abstraction needs to happen within the model itself, before the model can even attempt to intelligently manipulate it.

[fumeux_fume]

For many people, the difference between how a language model solves a problem and how a human solves a problem is actually very important.

[pessimizer]

> Pattern matching questions on a contrived test is not the same thing as understanding or reasoning.

Pattern matching is definitely the same thing as understanding and reasoning.

The problem is that LLMs can't recognize patterns that are longer than a few paragraphs, because the tokens would have to be far too long. LLMs are a thing we are lucky to have because we have very fast computers and very smart mathematicians making very hard calculations very efficient and parallelizable. But they sit on top of a bed of an enormous amount of human written knowledge, and can only stretch so far from that bed before completely falling apart.

Humans don't use tokenizers.

The goal right now is to build a scaffolding of these dummies in order to get really complicated work done, but that work is only ever going to accidentally be correct because of an accumulation of errors. This may be enough for a lot if we try it 1000x and run manually-tuned algos over the output to find the good ones. But this is essentially manual work, done in the traditional way.

edit: sorry, you're never going to convince me these things are geniuses when I chat to them for a couple of back and forth exchanges and they're already obviously losing track of everything, even what they just said. The good thing is that what they are is enough to do a lot, if you're a person who can be satisfied that they're not going to be your god anytime soon.