[HN Gopher] Hierarchical Reasoning Model
       ___________________________________________________________________
        
       Hierarchical Reasoning Model
        
       Author : hansmayer
       Score  : 275 points
       Date   : 2025-07-27 07:15 UTC (15 hours ago)
        
 (HTM) web link (arxiv.org)
 (TXT) w3m dump (arxiv.org)
        
       | torginus wrote:
       | Is it just me or are symbolic (or as I like to call it 'video
       | game') AI is seeping back into AI?
        
         | taylorius wrote:
         | Perhaps so - but represented in a trainable, neural form. Very
         | exciting!
        
           | cornholio wrote:
           | Natural general intelligences sure seem to work this way.
        
           | marcosdumay wrote:
           | Training databases is much easier than training neural
           | networks.
        
         | bobosha wrote:
         | But symbolic != hierarchical
        
       | cs702 wrote:
       | Based on a quick first skim of the abstract and the introduction,
       | the results from hierarchical reasoning (HRM) models look
       | _incredible_ :
       | 
       | > Using only 1,000 input-output examples, without pre-training or
       | CoT supervision, HRM learns to solve problems that are
       | intractable for even the most advanced LLMs. For example, it
       | achieves near-perfect accuracy in complex Sudoku puzzles (Sudoku-
       | Extreme Full) and optimal pathfinding in 30x30 mazes, where
       | state-of-the-art CoT methods completely fail (0% accuracy). In
       | the Abstraction and Reasoning Corpus (ARC) AGI Challenge 27,28,29
       | - a benchmark of inductive reasoning - HRM, trained from scratch
       | with only the official dataset (~1000 examples), with only 27M
       | parameters and a 30x30 grid context (900 tokens), achieves a
       | performance of 40.3%, which substantially surpasses leading CoT-
       | based models like o3-mini-high (34.5%) and Claude 3.7 8K context
       | (21.2%), despite their considerably larger parameter sizes and
       | context lengths, as shown in Figure 1.
       | 
       | I'm going to read this carefully, in its entirety.
       | 
       | Thank you for sharing it on HN!
        
         | diwank wrote:
         | Exactly!
         | 
         | > It uses two interdependent recurrent modules: a *high-level
         | module* for abstract, slow planning and a *low-level module*
         | for rapid, detailed computations. This structure enables HRM to
         | achieve significant computational depth while maintaining
         | training stability and efficiency, even with minimal parameters
         | (27 million) and small datasets (~1,000 examples).
         | 
         | > HRM outperforms state-of-the-art CoT models on challenging
         | benchmarks like Sudoku-Extreme, Maze-Hard, and the Abstraction
         | and Reasoning Corpus (ARC-AGI), where CoT methods fail
         | entirely. For instance, it solves 96% of Sudoku puzzles and
         | achieves 40.3% accuracy on ARC-AGI-2, surpassing larger models
         | like Claude 3.7 and DeepSeek R1.
         | 
         | Erm what? How? Needs a computer and sitting down.
        
           | mkagenius wrote:
           | Is it talking about fine tuning existing models with 1000
           | examples to beat them in those tasks?
        
           | cs702 wrote:
           | Yeah, that was pretty much my reaction. I will need time on a
           | computer too.
           | 
           | The repo is at https://github.com/sapientinc/HRM .
           | 
           | I _love it_ when authors publish _working code_. It 's
           | usually a good sign. If the code does what the authors claim,
           | no one can argue with it!
        
             | diwank wrote:
             | Same! Guan's work on sample packing during finetuning has
             | become a staple. His openchat code is also super simple and
             | easy to understand.
        
         | lumost wrote:
         | I am extremely skeptical of a 27M parameter model being trained
         | "from scratch" on 1000 datapoints. I am likewise incredulous of
         | the lack of comparison with any other model which is trained
         | "from scratch" using their data preparation. Instead they
         | strictly compare with 3rd party LLMs which are massively more
         | general purpose and may not have any of those 1000 examples in
         | their training set.
         | 
         | This smells like some kind of overfit to me.
        
       | electroglyph wrote:
       | but does it scale?
        
       | lispitillo wrote:
       | I hope/fear this HRM model is going to be merged with MoE very
       | soon. Given the huge economic pressure to develop powerful LLMs I
       | think this can be done in just a month.
       | 
       | The paper seems to only study problems like sudoku solving, and
       | not question answering or other applications of LLMs. Furthermore
       | they omit a section for future applications or fusion with
       | current LLMs.
       | 
       | I think anyone working in this field can envision their
       | applications, but the details to have a MoE with an HRM model
       | could be their next paper.
       | 
       | I only skimmed the paper and I am not an expert, sure other
       | will/can explain why they don't discuss such a new structure.
       | Anyway, my post is just blissful ignorance over the complexity
       | involved and the impossible task to predict change.
       | 
       | Edit: A more general idea is that Mixture of Expert is related to
       | cluster of concepts and now we would have to consider a cluster
       | of concepts related by the time they take to be grasped, so in a
       | sense the model would have in latent space an estimation of the
       | depth, number of layers, and time required for each concept, just
       | like we adapt our reading style for a dense math book different
       | to a newspaper short story.
        
         | buster wrote:
         | must say I am suspicious in this regard, as they don't show
         | applications other than a Sudoku solver and don't discuss
         | downsides.
        
           | Oras wrote:
           | and the training was only on Sudoku. Which means they need to
           | train a small model for every problem that currently exists.
           | 
           | Back to ML models?
        
             | lispitillo wrote:
             | Not only on Sudoku, there is also maze solving and ARC-AGI.
        
             | JBits wrote:
             | I would assuming that training a LLM would be unfeasible
             | for a small research lab, so isn't tackling small problems
             | like this unavoidable? Given that current LLMs have clear
             | limitations, I can't think of anything better than
             | developing beter architectures on small test cases, then a
             | company can try scaling it later.
        
         | yorwba wrote:
         | This HRM is essentially purpose-designed for solving puzzles
         | with a small number of rules interacting in complex ways.
         | Because the number of rules is small, a small model can learn
         | them. Because the model is small, it can be run many times in a
         | loop to resolve all interactions.
         | 
         | In contrast, language modeling requires storing a large number
         | of arbitrary phrases and their relation to each other, so I
         | don't think you could ever get away with a similarly small
         | model. Fortunately, a comparatively small number of steps
         | typically seems to be enough to get decent results.
         | 
         | But if you tried to use an LLM-sized model in an HRM-style
         | loop, it would be dog slow, so I don't expect anyone to try it
         | anytime soon. Certainly not within a month.
         | 
         | Maybe you could have a hybrid where an LLM has a smaller HRM
         | bolted on to solve the occasional constraint-satisfaction task.
        
           | energy123 wrote:
           | What about many small HRM models that solve conceptually
           | distinct subtasks as determined and routed to by a master
           | model who then analyzes and aggregates the outputs, with all
           | of that learned during training.
        
           | marcosdumay wrote:
           | > In contrast, language modeling requires storing a large
           | number of arbitrary phrases and their relation to each other
           | 
           | A person has some ~10k word vocabulary, with words fitting
           | specific places in a really small set of rules. All combined,
           | we probably have something on the order of a few million
           | rules in a language.
           | 
           | What, yes, is larger than the thing in this paper can handle.
           | But is nowhere near as large as a problem that should require
           | something the size of a modern LLM to handle. So it's well
           | worth it to try to enlarge models with other architectures,
           | try hybrid models (note that this one is necessarily hybrid
           | already), and explore every other possibility out there.
        
       | OgsyedIE wrote:
       | Skimming this, there is no reason why a MoE LLM system (whether
       | autoregressive, diffusion, energy-based or mixed) couldn't be
       | given a nested architecture that duplicates the layout of a HRM.
       | Combining these in different ways should allow for some novel
       | benchmarks around efficiency and quality, which will be
       | interesting.
        
       | 0x000xca0xfe wrote:
       | Goodbye captchas I guess? Somehow they are still around.
        
       | topspin wrote:
       | > "After completing the T steps, the H-module incorporates the
       | sub-computation's outcome (the final state L) and performs its
       | own update. This H update establishes a fresh context for the
       | L-module, essentially "restarting" its computational path and
       | initiating a new convergence phase toward a different local
       | equilibrium."
       | 
       | So they let the low-level RNN bottom out, evaluate the output in
       | the high level module, and generate a new context for the low-
       | level RNN. Rinse, repeat. The low-level RNNs are iterating
       | backpropagation while the high-level is periodically kicking the
       | low-level RNNs to get better outputs. Loops within loops.
       | Composition.
       | 
       | Another interesting part:
       | 
       | > "Neuroscientific evidence shows that these cognitive modes
       | share overlapping neural circuits, particularly within regions
       | such as the prefrontal cortex and the default mode network. This
       | indicates that the brain dynamically modulates the "runtime" of
       | these circuits according to task complexity and potential
       | rewards.
       | 
       | > Inspired by the above mechanism, we incorporate an adaptive
       | halting strategy into HRM that enables `thinking, fast and slow'"
       | 
       | A scheduler that dynamically balances resources based on the
       | necessary depth of reasoning and the available data.
       | 
       | I love how this paper cites parallels with real brains
       | throughout. I believe AGI will be solved as the primitives we're
       | developing are composed to extreme complexity, utilizing many
       | cooperating, competing, communicating, concurrent, specialized
       | "modules." It is apparent to me that human brain must have this
       | complexity, because it's the only feasible way evolution had to
       | achieve cognition using slow, low power tissue.
        
         | username135 wrote:
         | As soon I read the hlm/llm split, it immediately reminded me of
         | the human brain.
        
           | esafak wrote:
           | Composition is the whole point of deep learning. Deep as in
           | multilayer, multilevel.
        
             | dbagr wrote:
             | You need recursion at some point: you can't account for all
             | possible scenarios of combinations, as you would need an
             | infinite number of layers.
        
               | crystal_revenge wrote:
               | > infinite number of layers
               | 
               | That's not as impossible as it seems, Gaussian Processes
               | are equivalent to a Neural Network with infinite hidden
               | units, and any multilayer NN can be approximated by one
               | with a single, larger layer of hidden units.
        
               | advael wrote:
               | I mean recurrence is an attempt to allow approximation of
               | recursive processes, no?
        
       | JonathanRaines wrote:
       | I advise scepticism.
       | 
       | This work does have some very interesting ideas, specifically
       | avoiding the costs of backpropagation through time.
       | 
       | However, it does not appear to have been peer reviewed.
       | 
       | The results section is odd. It does not include include details
       | of how they performed the assesments, and the only numerical
       | values are in the figure on the front page. The results for ARC2
       | are (contrary to that figure) not top of the leaderboard
       | (currently 19% compared to HRMs 5%
       | https://www.kaggle.com/competitions/arc-prize-2025/leaderboa...)
        
         | d4rkn0d3z wrote:
         | Skepticism is best expressed by repeating the experiment and
         | comparing results. I'm game and I have 10 days off work next
         | month. I wonder what can be had in terms of full source and
         | data, etc. from the authors?
        
           | PJones2000 wrote:
           | 24 hours on a 4070. Seems quite doable.
           | https://github.com/sapientinc/HRM
        
           | JonathanRaines wrote:
           | Nice! They provide trained checkpoints on their GitHub.
           | Repeating their results would be a good start.
           | https://github.com/sapientinc/HRM
        
         | cs702 wrote:
         | The authors' code is at https://github.com/sapientinc/HRM .
         | 
         | In fields like AI/ML, I'll take a preprint with _working code_
         | over peer-reviewed work without any code, _always_ , even when
         | the preprint isn't well edited.
         | 
         |  _Everyone everywhere_ can review a preprint and its published
         | code, instead of a tiny number of hand-chosen reviewers who are
         | often overworked, underpaid, and on tight schedules.
         | 
         | If the authors' claims hold up, the work will gain recognition.
         | If the claims don't hold up, the work will eventually be
         | ignored. Credentials are basically irrelevant.
         | 
         | Think of it as open-source, distributed, global review. It may
         | be messy and ad-hoc, since no one is in charge, but it works
         | much better than traditional peer review!
        
           | smokel wrote:
           | I sympathize partially with your views, but how would this
           | work in practice? Where would the review comments be stored?
           | Is one supposed to browse Hacker News to check the validity
           | of a paper?
           | 
           | If a professional reviewer spots a serious problem, the paper
           | will not make it to a conference or journal, saving us a lot
           | of trouble.
        
             | belter wrote:
             | > If a professional reviewer spots a serious problem
             | 
             | Did that ever happen? :-)
        
               | atq2119 wrote:
               | Of course. As usual, you tend to not hear about it when a
               | system we rely on works well.
        
             | yorwba wrote:
             | Peer review is a way to distribute the work of identifying
             | which papers are potentially worth reading. If you're
             | starting from an individual paper and then ask yourself
             | whether it was peer reviewed or not, you're doing it wrong.
             | If you really need to know, read it yourself and accept
             | that you might just be wasting your time.
             | 
             | If you want to mostly read papers that have already been
             | reviewed, start with people or organizations you trust to
             | review papers in an area you're interested in and read what
             | they recommend. That could be on a personal blog or through
             | publishing a traditional journal, the difference doesn't
             | matter much.
        
               | conception wrote:
               | "Find papers that support what you want via online echo
               | chambers" isn't the advice you want to be giving but it
               | is the net result of it. Society needs trusted
               | institutions. Not that publishers are the best result of
               | that but adhoc blog posts are decidedly not better.
        
               | yorwba wrote:
               | It is totally the advice I want to be giving. Given the
               | choice between an echo chamber matched to my interests
               | and wading through a stream of unfiltered crap, I'll take
               | the echo chamber every time. (Of course there's also the
               | option of not reading papers at all, which is typically a
               | good choice if you're not a subject matter expert and
               | don't intend to put in the work to become one.)
               | 
               | If you choose to focus on the output of a well-known
               | publisher, you're not avoiding echo chambers, you're
               | using a heuristic to hopefully identify a good one.
        
               | ricardobeat wrote:
               | Those are not the only options, namely the parent
               | mentioned 'trusted institutions'. It is the best way to
               | defer that filtering to a group of other humans, whose
               | collective expertise will surpass any one individual.
               | 
               | The destruction of trust in both public and private
               | institutions - newspapers, journals, research
               | institutions, universities - and replacement with social
               | media 'influencers' and online echo chambers is how we
               | arrived at the current chaotic state of politics
               | worldwide, the rise of extremist groups, cults, a
               | resurgence of nationalism, religious fanaticism... This
               | is terrible advice.
        
         | sigmoid10 wrote:
         | Skepticism is an understatement. There are tons of issues with
         | this paper. Why are they comparing results of their expert
         | model that was trained from scratch on a single task to general
         | purpose reasoning models? It is well established in the
         | literature that you can still beat general purpose LLMs in
         | narrow domain tasks with specially trained, small models. The
         | only comparison that would have made sense is one to vanilla
         | transformers using the same nr of parameters and trained on the
         | same input-output dataset. But the paper shows no such
         | comparison. In fact, I would be surprised if it was
         | significantly better, because such architecture improvements
         | are usually very modest or not applicable in general. And
         | insinuating that this is some significant development to
         | improve general purpose AI by throwing in ARC is just straight
         | up dishonest. I could probably cook up a neural net in pytorch
         | in a few minutes that beats a hand-crafted single task that o3
         | can't solve in an hour. That doesn't mean that I made any
         | progress towards AGI.
        
           | bubblyworld wrote:
           | Have you spent much time with the ARC-1 challenge? Their
           | results on that are extremely compelling, showing results
           | close to the initial competition's SOTA (as of closing
           | anyway) with a tiny model and no hacks like data
           | augmentation, pretraining, etc that _all_ of the winning
           | approaches leaned on heavily.
           | 
           | Your criticism makes sense for the maze solving and sudoku
           | sets, of course, but I think it kinda misses the point (there
           | are traditional algos that solve those just fine - it's more
           | about the ability of neural nets to figure them out during
           | training, and known issues with existing recurrent
           | architectures).
           | 
           | Assuming this isn't fake news lol.
        
             | smokel wrote:
             | Looking at the code, there is a lot of data augmentation
             | going on there. For the Sudoku and ARC data sets, they
             | augment every example by a factor of 1,000.
             | 
             | https://github.com/sapientinc/HRM/blob/main/dataset/build_a
             | r...
        
               | bubblyworld wrote:
               | That's fair, they are relabelling colours and rotating
               | the boards. I meant more like mass generation of novel
               | puzzles to try and train specific patterns. But you are
               | right that technically there is some augmentation going
               | on here, my bad.
        
               | smokel wrote:
               | Hm, I'm not so sure it's fair play for the Sudoku puzzle.
               | Suggesting that the AI will understand the rules of the
               | game with only 1,000 examples, and then adding 1,000,000
               | derived examples does not feel fair to me. Those extra
               | examples leak a lot of information about the rules of the
               | game.
               | 
               | I'm not too familiar with the ARC data set, so I can't
               | comment on that.
        
               | bubblyworld wrote:
               | True, it leaks information about all the symmetries of
               | the puzzle, but that's about it. I guess someone needs to
               | test how much that actually helps - if I get the model
               | running I'll give it a try!
        
               | westurner wrote:
               | > _That 's fair, they are relabelling colours and
               | rotating the boards._
               | 
               | Photometric augmentation, Geometric augmentation
               | 
               | > _I meant more like mass generation of novel puzzles to
               | try and train specific patterns._
               | 
               | What is the difference between Synthetic Data Generation
               | and Self Play (like AlphaZero)? Don't self play
               | simulations generate _synthetic_ training data as
               | compared to real observations?
        
               | bubblyworld wrote:
               | I don't know the jargon, but for me the main thing is the
               | distinction between humans injecting additional bits of
               | information into the training set vs the algorithm itself
               | discovering those bits of information. So self-play is
               | very interesting (it's automated as part of the
               | algorithm) but stuff like generating tons of novel sudoku
               | puzzles and adding them to the training set is less
               | interesting (the information is being fed into the
               | training set "out-of-band", so to speak).
               | 
               | In this case I was wrong, the authors are clearly adding
               | bits of information themselves by augmenting the dataset
               | with symmetries (I propose "symmetry augmentation" as a
               | much more sensible phrase for this =P). Since symmetries
               | share a lot of mutual information with each other, I
               | don't think this is nearly as much of a crutch as adding
               | novel data points into the mix before training, but
               | ideally no augmentation would be needed.
               | 
               | I guess you could argue that in some sense it's fair play
               | - when humans are told the rules of sudoku the symmetry
               | is implicit, but here the AI is only really "aware" of
               | the gradient.
        
             | sigmoid10 wrote:
             | As the other commenter already pointed out, I'll believe it
             | when I see it on the leaderboard. But even then it already
             | lost twice against the winner of last year's competition,
             | because that too was a general purpose LLM that could also
             | do other things.
        
               | bubblyworld wrote:
               | Let's not move the goalposts here =) I don't think it's
               | really fair to compare them directly like that. But I
               | agree, this is triggering my "too good to be true" reflex
               | very hard.
        
               | sigmoid10 wrote:
               | If anything, they moved the goalpost closer to the
               | starting line. I'm merely putting it back where it
               | belongs.
        
         | diwank wrote:
         | I think that's too harsh a position solely for not being peer
         | reviewed yet. Neither of yhe original mamba1 and mamba2 papers
         | were peer reviewed. That said, strong claims warrant strong
         | proofs, and I'm also trying to reproduce the results locally.
        
         | hodgehog11 wrote:
         | Scepticism is generally always a good idea with ML papers. Once
         | you start publishing regularly in ML conferences, you
         | understand that there is no traditional form of peer review
         | anymore in this domain. The volume of papers has meant that
         | 'peers' are often students coming to grips with parts of the
         | field that rarely align with what they are asked to review.
         | Conference peer review has become a 'vibe check' more than
         | anything.
         | 
         | Real peer review is when other experts independently verify
         | your claims in the arXiv submission through implementation and
         | (hopefully) cite you in their followup work. This thread is
         | real peer review.
        
           | dleeftink wrote:
           | I appreciate this insight, makes you wonder, why even publish
           | a paper if it only amounts to a vibe check? If it's just the
           | code we need we can get that peer reviewed through other
           | channels.
        
             | thfuran wrote:
             | Because publications is the number that academics have to
             | make go up.
        
               | hodgehog11 wrote:
               | This and the exposure. There are so many papers on arXiv
               | now that people often look to conference or journal
               | publication lists.
        
               | dleeftink wrote:
               | The number has clearly ceased its function, so what are
               | we chasing?
        
               | gavinray wrote:
               | Clout, funding, and employment I'd imagine?
        
           | rapatel0 wrote:
           | THIS is so true but also not limited to ML.
           | 
           | Having been both a publisher and reviewer across multiple
           | engineering, science, and bio-medical disciplines this occurs
           | across academia.
        
         | frozenseven wrote:
         | >does not appear to have been peer reviewed
         | 
         | Enough already. Please. The paper + code is here for everybody
         | to read and test. Either it works or it doesn't. Either people
         | will build upon it or they won't. I don't need to wait 20
         | months for 3 anonymous dudes to figure it out.
        
         | riku_iki wrote:
         | > However, it does not appear to have been peer reviewed.
         | 
         | my observation is that peer reviewers never try to reproduce
         | results or do basic code audit to check that there is no data
         | leak for example to training dataset.
        
         | mitthrowaway2 wrote:
         | Do you consider yourself a peer? Feel free to review it.
         | 
         | A peer reviewer will typically comment that some figures are
         | unclear, that a few relevant prior works have gone uncited, or
         | point out a followup experiment that they should do.
         | 
         | That's about the extent of what peer reviewers do, and
         | basically what you did yourself.
        
         | halayli wrote:
         | The fact that you are expecting a paper just published to have
         | been peer reviewed already tells me that you are likely not
         | familiar with the process. The first step to have your work
         | peer reviewed is to publish it.
        
       | smokel wrote:
       | If I understand this correctly, it learns the _rules_ of Sudoku
       | by looking at 1,000 examples of (puzzle, solution) pairs. It is
       | then able to solve previously unseen puzzles with 55% accuracy.
       | If given millions of examples, it becomes almost perfect.
       | 
       | This is apparently without pretraining of any sort, which is kind
       | of amazing. In contrast, systems like AlphaZero have the rules to
       | go or chess built-in, and only learn the strategy, not the rules.
       | 
       | Off to their GitHub repository [1] to see this for myself.
       | 
       | [1] https://github.com/sapientinc/HRM
        
         | babel_ wrote:
         | AlphaZero may have the rules built in, but MuZero and the other
         | follow-ups didn't. MuZero not only matched or surpassed
         | AlphaZero, but it did so with less training, especially in the
         | EfficientZero variant; notably also on the Atari playground.
        
           | smokel wrote:
           | Thanks for pointing that out.
           | 
           | To be fair, MuZero only learns a model of the rules for
           | navigating its search tree. To make actual moves, it gets a
           | list of valid actions from the game engine, so at that level
           | it does not learn the rules of the game.
           | 
           | (HRM possibly does the same, and could be in the same realm
           | as MuZero. It probably makes a lot of illegal moves.)
        
           | gavmor wrote:
           | This is "The Bitter Lesson" of AI, no? "More compute beats
           | clever algorithm."
        
             | babel_ wrote:
             | Quite the opposite, a clever algorithm needs less compute,
             | and can leverage extra compute even more.
        
               | gavmor wrote:
               | Apologies, "clever" is a poor paraphrase of "domain-
               | specific", or "methods that leveraged human
               | understanding."[0]
               | 
               | 0.
               | http://www.incompleteideas.net/IncIdeas/BitterLesson.html
        
         | smokel wrote:
         | To follow up, after experimenting a bit with the source code:
         | 
         | 1. Please, for the love of God, and for scientific
         | reproducibility, specify library versions explicitly, and use
         | pyproject.toml instead of an incomplete requirements.txt.
         | 
         | 2. The 1,000 Sudoku examples are augmented with hand-coded
         | permutation algorithms, so the actual input data set is more
         | like 1,000,000 examples, not 1,000.
        
           | rudedogg wrote:
           | Do you have a fork or the changes? I might take a look, and
           | python dependency hell on Sunday is no good
        
           | mkagenius wrote:
           | > specify library versions explicitly
           | 
           | Sometimes even that is not helpful. It's a pain we have to
           | deal with.
        
             | gavinray wrote:
             | How is it not helpful?
             | 
             | A dependency lock file with resolved versions for both
             | direct and transient dependencies = reproducible build
        
               | mkagenius wrote:
               | I don't remember the exact scenario but it might have
               | been related to the underlying python or some sys library
               | being a little different and then the dependency lock not
               | being compatible with it.
        
               | blincoln wrote:
               | I don't know how common this is, but the fschat library
               | maintainers went for at least a year without making an
               | official release or updating the version number in their
               | GitHub repo, so the only way to both have current code
               | _and_ a reproducible build (without just including the
               | fschat library directly, of course) was to pin it to a
               | particular GitHub commit hash, which would get you code
               | that was current, but with the version number from 12+
               | months earlier.
               | 
               | fschat is pretty popular for LLM-related work, so I
               | assume this is at least not unheard-of for other notable
               | third-party libraries.
        
       | belter wrote:
       | Is this not a variation of ReAct + Chain-of-Thought + Structured
       | Planning? Or is that too unfair to the authors work?
       | 
       | [1] - https://arxiv.org/abs/2210.03629
        
       | SubiculumCode wrote:
       | As a cognitive psychologist, I highly suspected that, broadly
       | speaking, this was the needed direction for AI. See Fuzzy Trace
       | Theory[1].
       | 
       | Fuzzy Trace Theory basically suggests that memory (and cognition
       | generally) works at multiple levels spanning verbatim
       | representations to gist-level representations, that get bound
       | together into memories. Recalling gist, the general idea, along
       | with specific details, allows for powerful generalization and
       | flexible retrieval pathways.
       | 
       | [1] https://pmc.ncbi.nlm.nih.gov/articles/PMC4979567/
        
       | malcontented wrote:
       | I appreciate the connections with neurology, and the paper itself
       | doesn't ring any alarm bells. I don't think I'd reject it if it
       | fell to me to peer review.
       | 
       | However, I have extreme skepticism when it comes to the
       | applicability of this finding. Based on what they have written,
       | they seem to have created a universal (maybe; adaptable at the
       | very least) constraint-satisfaction solver that learns the rules
       | of the constraint-satisfaction problem from a small number of
       | examples. If true (I have not yet had the leisure to replicate
       | their examples and try them on something else), this is pretty
       | cool, but I do not understand the comparison with CoT models.
       | 
       | CoT models can, in principle, solve _any_ complex task. This
       | needs to be trained to a specific puzzle which it can then solve:
       | it makes no pretense to universality. It isn't even clear that it
       | is meant to be capable of adapting to any given puzzle. I suspect
       | this is not the case, just based on what I have read in the paper
       | and on the indicative choice of examples they tested it against.
       | 
       | This is kind of like claiming that Stockfish is way smarter than
       | current state of the art LLMs because it can beat the stuffing
       | out of them in chess.
       | 
       | I feel the authors have a good idea here, but that they have
       | marketed it a bit too... generously.
        
         | jurgenaut23 wrote:
         | Yes, I agree, but this is a huge deal in and of itself. I
         | suppose the authors had to frame it in this way for obvious
         | reasons of hype surfing, but this is an amazing achievement,
         | especially given the small size of the model! I'd rather use a
         | customized model for a specific problem than a supposedly <<
         | generally intelligent >> model that burns orders of magnitude
         | more energy for much less reliability.
        
         | JBits wrote:
         | > CoT models can, in principle, solve _any_ complex task.
         | 
         | What is the justification for this? Is there a mathematical
         | proof? To me, CoT seems like a hack to work around the severe
         | limitations of current LLMs.
        
           | liamnorm wrote:
           | The Universal Approximation Theorem.
        
             | JBits wrote:
             | I don't see how that changes anything. By this logic,
             | there's no need for CoT reasoning at all, as a single pass
             | should be sufficient. I don't see how that proves that CoT
             | increases capabilities.
        
       | advael wrote:
       | I really like this usage of recurrent modules to augment
       | attention-based models, and I think this is a really cool result
       | and a fruitful avenue for future work
        
       | camphy wrote:
       | This is really interesting, but does anyone think this is
       | something that might generalize for ambiguous reasoning
       | situations with more development? I am no expert, but sudoku and
       | puzzles seem like very well-defined problem spaces.
        
       | nowittyusername wrote:
       | I've been keeping an eye on this one as well. based on what the
       | paper claims this would be huge. But i think like many here, we
       | are waiting for either confirmation or denial of the claim via 3d
       | parties. the concept behind them sounds legit, but id like to see
       | it in practice.
        
       ___________________________________________________________________
       (page generated 2025-07-27 23:00 UTC)