Rendered at 04:09:15 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
robot-wrangler 2 hours ago [-]
> Frontier models score ~90% on Python but only 3.8% on esoteric languages, exposing how current code generation relies on training data memorization rather than genuine programming reasoning.
Finally! This is a really obvious test-case that I've wondered about myself, and have seen many casual skeptics and cautiously optimistic people independently raising for several years now. When megacorp is not crowing about such a test, the silence is deafening, and it was practically guaranteed that they tested, didn't like the results, and didn't publish.
I'm still surprised it took this long for academics to try it, and skimming cites, I don't see anything similar. Anyone know if this is the first paper to try this kind of thing, or just the first paper to put together a especially good suite of reusable benchies?
If this benchmark becomes popular, then presumably to avoid such embarrassments synthetic data is eventually added to training sets to make sure even esolangs are somewhat more in-distro, and then we gradually run out of esolangs to do honest testing with. SAT is a whole different animal admittedly, but comparable honest tests might involve just forcing models to use randomly generated but easily checked EBNF grammar? I don't have a quick link to the relevant papers, but afaik benchmarks of strict adherence to non-simple JSON schemas is also still pretty bad, and we're just working around it with lots of retries/tokens. "But look how well it works for 10k lines of kubernetes manifests!" Well yeah, maybe, but it barely needs to really follow a schema since that is more stuff that's in the training set..
chromaton 53 minutes ago [-]
I did something very similar last year, but with programming languages that were REALLY out of distribution; they were generated specifically for the benchmark. I call it TiānshūBench (天书Bench): https://jeepytea.github.io/general/introduction/2025/05/29/t...
Some models were OK at solving very simple problems, but nearly all of them would, for example, hallucinate control structures that did not exist in the target language.
amluto 10 minutes ago [-]
A few months ago I created a little API to help with an obnoxious case in FFI [0] in an extremely esoteric language known as Python. It was straightforward, it had a fully typed signature, and I fully documented it. (And the entire implementation was only 50 lines or so of intentionally very straightforward code.) The LLM (Codex 5.2 IIRC) could not manage to call the function with the right arguments even after multiple rounds of prompting.
Sometimes I think LLMs are unbelievably, amazingly good at things. And sometimes I’m deeply suspicious that they really not very smart, and this was an example of the latter.
[0] Python calling to C, passing a callback function pointer and a void *opaque that C will pass back to the callback. Short of writing an extension module, this is pretty much forced to go through an inherently nasty JIT codegen process in libffi, which is sort of tolerable, but you really don’t want to redo it for each object that gets opacified to void*. Codex passed a lambda, which did the nasty JIT thing every time. I wrote a little shim using weakref. Apparently no one has done this before, so Codex wasn’t trained on it, and it couldn’t make itself call the function. Maybe I should post it to PyPI.
orthoxerox 6 hours ago [-]
> Frontier models score ~90% on Python but only 3.8% on esoteric languages, exposing how current code generation relies on training data memorization rather than genuine programming reasoning.
I would probably score about the same, does this prove I also rely on training data memorization rather than genuine programming reasoning?
Or does this simply show that esolangs are hard to reason in by design? A more honest approach would use a "real", but relatively unpopular, language. Make them use CoffeeScript or Ada or PL/I or Odin or that other systems programming language that that very opinionated guy is implementing on top of QBE.
onoesworkacct 5 hours ago [-]
Unlike AI, you aren't able to regurgitate entire programs and patterns you've seen before.
AI's capacity for memorisation is unrivaled, I find it mind blowing that you can download a tiny ~4gb model and it will have vastly more general knowledge than an average human (considering that the human is more likely to be wrong if you ask it trivia about e.g. the spanish civil war).
But the average human still has actual reasoning capabilities, which is still (I think?) a debated point with AI.
refulgentis 3 hours ago [-]
> which is still (I think?) a debated point with AI.
It's not, people misread an Apple study and it became a meme. It lost currency as a meme because it is impossible to use a model in 2026 and come away with the idea it cannot reason, for any reasonable definition of the word reason (pun intended). Most of the debate from there is just people misreading each-other and imagining incentive structures at play. (ex. I am not claiming they are never stupid, ex. the car wash dilemma, but I am claiming its gee-whiz enough at enough that it's become de facto beyond honest debate)
> AI's capacity for memorisation is unrivaled,
Much like "it just memorizes training data", "memorization" has a kernel of truth to it. Memorizing does not imply "it has 100% "learned", for some definition of learned similar to "guaranteed 100% reproducible translatable computation", brainfuck to the point it's just as easy as writing any other program, and thus if it hasn't, it cannot reason"
At the end of the day these are just mathematical objects. And while it's not discourse-contributing, the mundane truth is, those matmuls born from boring curve-fitting at scale know/memorized/can reason about/can parrot/have adjusted the float32s in such a way that it produces C a lot better than Brainfuck. Much like us. But they're just matmuls curve-fitting at scale.
IsTom 5 hours ago [-]
Just look what kind of problems the easy task set is (hello world, echo line, count vowels, etc.). With best being ~10% of total in brainfuck this is 10 out of 20. You can google more solutions to these problems than that.
voxl 4 hours ago [-]
It's pointless to argue, we exist in world of "this technology will usher in the singularity" versus "this tech is useful but come on"
The singularity crowd has never listened to reason and never will.
andai 5 hours ago [-]
Yeah there seem to be two axes here.
Esolang vs mainstream paradigm.
Popular vs scarce training data.
So you'd want to control for training data (e.g. brainfuck vs Odin?)
And ideally you'd control by getting it down to 0, i.e. inventing new programming languages with various properties and testing the LLMs on those.
I think that would be a useful benchmark for other reasons. It would measure the LLMs' ability to "learn" on the spot. From what I understand, this remains an underdeveloped area of their intelligence. (And may not be solvable with current architectures.)
astrange 4 hours ago [-]
> I would probably score about the same, does this prove I also rely on training data memorization rather than genuine programming reasoning?
It doesn't even prove the models do that. The RLVR environments being mostly Python isn't "training data memorization". That's just the kind of dumb thing people say to sound savvy.
Pamar 2 hours ago [-]
I had similar experiences with an unpopular but not "esoteric" language (Progress ABL) and so did some other developers in my team.
derrak 2 hours ago [-]
I don’t know your background, but suspect that if you were given sufficient motivation, you could solve these problems in an esoteroic language. It might be tedious, but I suspect that most anyone with an undergraduate degree in computer science and sufficient experience in a couple programming languages could meet the task.
iloveoof 5 hours ago [-]
Try MUMPS, widely used but little training data online. Probably less than some esolangs
twoodfin 4 hours ago [-]
Frontier models have gotten much better at ObjectScript (the InterSystems evolution of MUMPS/M).
I wish it would use Return instead of Quit but that’s a stochastic parrot for you.
wavemode 5 hours ago [-]
> I would probably score about the same, does this prove I also rely on training data memorization rather than genuine programming reasoning?
Setting aside whether this benchmark is meaningful or not - the argument you're making is faulty. There are indeed humans who can write complete programs in Brainfuck and these other esolangs. The fact that you personally can't is not logically relevant.
Groxx 5 hours ago [-]
particularly if you'd already read approximately all written material in existence about those languages. many humans are capable of learning a language from the documentation.
ventisk1ze 2 hours ago [-]
[dead]
bwestergard 7 hours ago [-]
I'm shocked to see how poorly these models, which I find useful day to day, do in solving virtually any of the problems in Unlambda.
Before looking at the results my guess was that scores would be higher for Unlambda than any of the others, because humans that learn Scheme don't find it all that hard to learn about the lambda calculus and combinatory logic.
But the model that did the best, Qwen-235B, got virtually every problem wrong.
__alexs 7 hours ago [-]
They are also weirdly bad at Brainfuck which is basically just a subset of C.
astrange 4 hours ago [-]
BF involves a lot of repeated symbols, which is hard for tokenized models. Same problem as r's in strawberry.
bwestergard 4 hours ago [-]
Interesting. So why do the models seem to handle deeply nested Lisp expressions just fine?
kgeist 3 hours ago [-]
Probably because there's a ton of code that deals with nested parentheses across languages in the training data, and models have learned how to work around tokenization limitations, when it comes to parentheses.
paraschopra 1 hours ago [-]
(founder of Lossfunk, the lab behind this research.)
Esolang-Bench went viral on X. A lot of discussion ensued; addressing some of the common points that came up. Addressing a few questions about our Esolang-Bench. Hope it helps.
a) Why do it? Does it measure anything useful?
It was a curiosity-driven project. We're interested in how humans exhibit sample-efficiency in learning and OOD generalization. So we simply asked: if models can zero/few shot correct answers for simple programming problems in Python, can they do the same in esoteric languages as well?
The benchmark is what it is. Different people can interpret its usefulness differently, and we encourage that.
b) But humans can't also write esoteric languages well. It's an unfair comparison.
Primarily, we're interested in measuring LLM capabilities. With the talk of ASI, it is supposed that their capabilities will soon be super-human. So, our primary motivation wasn't to compare to humans but to check what they can do this by-construction difficult benchmark.
However, we do believe that humans are able to teach themselves a new domain by transferring their old skills. So this benchmark was to set a starting point to explore how AI systems can do the same as well (which is what we're exploring now)
c) But Claude Code crushes it. You limited models artificially.
Yes, we tested models in zero and few shot capabilities. And in the agentic loop we describe in the paper, we limit the number of iterations. As we wrote above, we wanted to understand their performance from a comparative point of view (say on highly represented languages like Python) and that's by the benchmark by design is like this.
After the paper was finalized, we experimented with agentic systems where we gave models tools like bash and allowed unlimited iterations (but limited submission attempts). They indeed perform much better.
The question that's relevant is what makes these models perform so well when you give them tools and iterations v/s when you don't. Are they reasoning / learning like humans or is it something else?
d) So, are LLMs hyped? Or is our study clickbait?
The paper, code and benchmark are all open source.
We encourage whoever is interested to read it, and make up their own minds.
(We couldn't help notice that the same set of results were interpreted wildly differently within the community. A debate between opposing camps of LLMs ensued. Perhaps that's a good thing?)
monster_truck 4 hours ago [-]
I have encountered the opposite of this. All of the latest pro tier models are still fighting for their lives to use powershell correctly, really basic things like quotes, escaping, heredocs. Doesn't matter what I put in agents.md or instruct it to do. You just have to accept the token tax of it stomping on rakes until it figures it out itself and then keep using that session.
It's bad enough that I've considered writing some sort of cursed bash->posh translation layer
Yet it has no issues at all implementing and then writing slopjective-c 3.0
noahbp 3 hours ago [-]
Opus 4.6 has gotten pretty good at writing Powershell.
It’s the first model where I didn’t have to ask, repeatedly, that it use Powershell 5, and never use emojis or other invalid characters, like Gemini and those non-ASCII spaces.
__alexs 7 hours ago [-]
I had hope we might finally be ushering in a bold new era of programming in Malbolge but apparently that was too optimistic.
sathish316 3 hours ago [-]
Does this imply LLMs will not work well on novel reasoning problems?
danpalmer 56 minutes ago [-]
Yep that's the implication. Anecdotally this is obvious to me. I'm using LLMs to write Java and C++, and then can churn out generic plumbing with no issues, but novel code for a novel implementation of a novel idea, they have no idea what they're doing.
I'm getting good productivity gains, but it requires a lot of hand holding because AI does not know what it's doing.
On far less novel problems I get far better results.
wmf 3 hours ago [-]
ARC-AGI is already testing that.
groar 4 hours ago [-]
I guess if you tell codex to build a transpiler from a subset of python to brainfuck, then solve in that subset of python, it would work much better. Would that be cheating?
simianwords 7 hours ago [-]
I bet I can do better by allowing this: the llm can pull documentation of the language from the web to understand how it works.
If the llm has “skills” for that language, it will definitely increase accuracy.
deklesen 7 hours ago [-]
Mhh... my hunch is that part of this is that all python keywords are 1 token, I assume. And for those very weird languages, tokenizing might make it harder to reason over those tokens.
Would love to see how the benchmarks results change if the esoteric languages are changed a bit to make them have 1-token keywords only.
chychiu 7 hours ago [-]
Considering that brainfuck only has 8 characters and models are scoring at 6.2% I don't think tokenization is the issue
altruios 6 hours ago [-]
The only issue. *
Reasoning is hard, reasoning about colors while wearing glasses that obfuscate the real colors... even harder... but not the core issue if your brain not wired correctly to reason.
I suspect the way out of this is to separate knowledge from reason: to train reasoning with zero knowledge and zero language... and then to train language on top of a pre-trained-for-reasoning model.
onoesworkacct 4 hours ago [-]
LLMs already use mixture of experts models, if you ensure the neurons are all glued together then (i think) you train language and reason simultaneously
rubyn00bie 4 hours ago [-]
I am not surprised by this, and am glad to see a test like this. One thing that keeps popping up for me when using LLMs is the lack of actual understanding. I write Elixir primarily and I can say without a doubt, that none of the frontier models understand concurrency in OTP/Beam. They look like they do, but they’ll often resort to weird code that doesn’t understand how “actors” work. It’s an imitation of understanding that is averaging all the concurrency code it has seen in training. With the end result being huge amount of noise, when those averages aren’t enough, guarding against things that won’t happen, because they can’t… or they actively introduce race conditions because they don’t understand how message passing works.
Current frontier models are really good at generating boiler plate, and really good at summarizing, but really lack the ability to actually comprehend and reason about what’s going on. I think this sort of test really highlights that. And is a nice reminder that, the LLMs, are only as good as their training data.
When an LLM or some other kind of model does start to score well on tests like this, I’d expect to see better them discovering new results, solutions, and approaches to questions/problems. Compared to how they work now, where they generally only seem to uncover answers that have been obfuscated but are present.
Finally! This is a really obvious test-case that I've wondered about myself, and have seen many casual skeptics and cautiously optimistic people independently raising for several years now. When megacorp is not crowing about such a test, the silence is deafening, and it was practically guaranteed that they tested, didn't like the results, and didn't publish.
I'm still surprised it took this long for academics to try it, and skimming cites, I don't see anything similar. Anyone know if this is the first paper to try this kind of thing, or just the first paper to put together a especially good suite of reusable benchies?
If this benchmark becomes popular, then presumably to avoid such embarrassments synthetic data is eventually added to training sets to make sure even esolangs are somewhat more in-distro, and then we gradually run out of esolangs to do honest testing with. SAT is a whole different animal admittedly, but comparable honest tests might involve just forcing models to use randomly generated but easily checked EBNF grammar? I don't have a quick link to the relevant papers, but afaik benchmarks of strict adherence to non-simple JSON schemas is also still pretty bad, and we're just working around it with lots of retries/tokens. "But look how well it works for 10k lines of kubernetes manifests!" Well yeah, maybe, but it barely needs to really follow a schema since that is more stuff that's in the training set..
Some models were OK at solving very simple problems, but nearly all of them would, for example, hallucinate control structures that did not exist in the target language.
Sometimes I think LLMs are unbelievably, amazingly good at things. And sometimes I’m deeply suspicious that they really not very smart, and this was an example of the latter.
[0] Python calling to C, passing a callback function pointer and a void *opaque that C will pass back to the callback. Short of writing an extension module, this is pretty much forced to go through an inherently nasty JIT codegen process in libffi, which is sort of tolerable, but you really don’t want to redo it for each object that gets opacified to void*. Codex passed a lambda, which did the nasty JIT thing every time. I wrote a little shim using weakref. Apparently no one has done this before, so Codex wasn’t trained on it, and it couldn’t make itself call the function. Maybe I should post it to PyPI.
I would probably score about the same, does this prove I also rely on training data memorization rather than genuine programming reasoning?
Or does this simply show that esolangs are hard to reason in by design? A more honest approach would use a "real", but relatively unpopular, language. Make them use CoffeeScript or Ada or PL/I or Odin or that other systems programming language that that very opinionated guy is implementing on top of QBE.
AI's capacity for memorisation is unrivaled, I find it mind blowing that you can download a tiny ~4gb model and it will have vastly more general knowledge than an average human (considering that the human is more likely to be wrong if you ask it trivia about e.g. the spanish civil war).
But the average human still has actual reasoning capabilities, which is still (I think?) a debated point with AI.
It's not, people misread an Apple study and it became a meme. It lost currency as a meme because it is impossible to use a model in 2026 and come away with the idea it cannot reason, for any reasonable definition of the word reason (pun intended). Most of the debate from there is just people misreading each-other and imagining incentive structures at play. (ex. I am not claiming they are never stupid, ex. the car wash dilemma, but I am claiming its gee-whiz enough at enough that it's become de facto beyond honest debate)
> AI's capacity for memorisation is unrivaled,
Much like "it just memorizes training data", "memorization" has a kernel of truth to it. Memorizing does not imply "it has 100% "learned", for some definition of learned similar to "guaranteed 100% reproducible translatable computation", brainfuck to the point it's just as easy as writing any other program, and thus if it hasn't, it cannot reason"
At the end of the day these are just mathematical objects. And while it's not discourse-contributing, the mundane truth is, those matmuls born from boring curve-fitting at scale know/memorized/can reason about/can parrot/have adjusted the float32s in such a way that it produces C a lot better than Brainfuck. Much like us. But they're just matmuls curve-fitting at scale.
The singularity crowd has never listened to reason and never will.
Esolang vs mainstream paradigm.
Popular vs scarce training data.
So you'd want to control for training data (e.g. brainfuck vs Odin?)
And ideally you'd control by getting it down to 0, i.e. inventing new programming languages with various properties and testing the LLMs on those.
I think that would be a useful benchmark for other reasons. It would measure the LLMs' ability to "learn" on the spot. From what I understand, this remains an underdeveloped area of their intelligence. (And may not be solvable with current architectures.)
It doesn't even prove the models do that. The RLVR environments being mostly Python isn't "training data memorization". That's just the kind of dumb thing people say to sound savvy.
Palindrome:
https://chatgpt.com/s/t_69bc8d8c116c8191a339a33f0fbcc935
This is a noticeable improvement from a year ago.
I wish it would use Return instead of Quit but that’s a stochastic parrot for you.
Setting aside whether this benchmark is meaningful or not - the argument you're making is faulty. There are indeed humans who can write complete programs in Brainfuck and these other esolangs. The fact that you personally can't is not logically relevant.
Before looking at the results my guess was that scores would be higher for Unlambda than any of the others, because humans that learn Scheme don't find it all that hard to learn about the lambda calculus and combinatory logic.
But the model that did the best, Qwen-235B, got virtually every problem wrong.
Esolang-Bench went viral on X. A lot of discussion ensued; addressing some of the common points that came up. Addressing a few questions about our Esolang-Bench. Hope it helps.
a) Why do it? Does it measure anything useful?
It was a curiosity-driven project. We're interested in how humans exhibit sample-efficiency in learning and OOD generalization. So we simply asked: if models can zero/few shot correct answers for simple programming problems in Python, can they do the same in esoteric languages as well?
The benchmark is what it is. Different people can interpret its usefulness differently, and we encourage that.
b) But humans can't also write esoteric languages well. It's an unfair comparison.
Primarily, we're interested in measuring LLM capabilities. With the talk of ASI, it is supposed that their capabilities will soon be super-human. So, our primary motivation wasn't to compare to humans but to check what they can do this by-construction difficult benchmark.
However, we do believe that humans are able to teach themselves a new domain by transferring their old skills. So this benchmark was to set a starting point to explore how AI systems can do the same as well (which is what we're exploring now)
c) But Claude Code crushes it. You limited models artificially.
Yes, we tested models in zero and few shot capabilities. And in the agentic loop we describe in the paper, we limit the number of iterations. As we wrote above, we wanted to understand their performance from a comparative point of view (say on highly represented languages like Python) and that's by the benchmark by design is like this.
After the paper was finalized, we experimented with agentic systems where we gave models tools like bash and allowed unlimited iterations (but limited submission attempts). They indeed perform much better.
The question that's relevant is what makes these models perform so well when you give them tools and iterations v/s when you don't. Are they reasoning / learning like humans or is it something else?
d) So, are LLMs hyped? Or is our study clickbait?
The paper, code and benchmark are all open source.
We encourage whoever is interested to read it, and make up their own minds.
(We couldn't help notice that the same set of results were interpreted wildly differently within the community. A debate between opposing camps of LLMs ensued. Perhaps that's a good thing?)
It's bad enough that I've considered writing some sort of cursed bash->posh translation layer
Yet it has no issues at all implementing and then writing slopjective-c 3.0
It’s the first model where I didn’t have to ask, repeatedly, that it use Powershell 5, and never use emojis or other invalid characters, like Gemini and those non-ASCII spaces.
I'm getting good productivity gains, but it requires a lot of hand holding because AI does not know what it's doing.
On far less novel problems I get far better results.
If the llm has “skills” for that language, it will definitely increase accuracy.
Would love to see how the benchmarks results change if the esoteric languages are changed a bit to make them have 1-token keywords only.
Reasoning is hard, reasoning about colors while wearing glasses that obfuscate the real colors... even harder... but not the core issue if your brain not wired correctly to reason.
I suspect the way out of this is to separate knowledge from reason: to train reasoning with zero knowledge and zero language... and then to train language on top of a pre-trained-for-reasoning model.
Current frontier models are really good at generating boiler plate, and really good at summarizing, but really lack the ability to actually comprehend and reason about what’s going on. I think this sort of test really highlights that. And is a nice reminder that, the LLMs, are only as good as their training data.
When an LLM or some other kind of model does start to score well on tests like this, I’d expect to see better them discovering new results, solutions, and approaches to questions/problems. Compared to how they work now, where they generally only seem to uncover answers that have been obfuscated but are present.