Okay, we're live.
We have one person.
People will start trickling in.
Thanks for coming to Sail Live number six.
This is a very exciting one.
I think we have a, I mean, the topics are always fun with these.
Whatever is the topic of the day on our little rat racing minds trying to keep up with AI.
But we're welcoming the latest writer that is joining the SAIL Coalition.
So I think this just means more content for SAIL.
I think I've been a fan of Swix and a friend for a while at this point.
So I'm very happy to have his content join this.
And I think you've been doing great stuff recently and continuing to evolve this.
Thank you, sir.
Welcome to the team.
This is my friends and colleagues in the AI media space and it's just great to be able to support people and keep that network closer.
Thanks for, I just want to say thanks for joining us.
It's really a pleasure to have you on here, Sean or Swix.
So yeah, awesome.
I just coincidentally listened to your podcast about the super benchmark.
So yeah, awesome to, you know, small world.
Awesome to have you here.
Yeah, thanks for having me.
And yeah, just glad to be on in chat.
I've never ever done one of these Substack live things.
So I'm curious how it works.
Because I always think about Substack because it's like a newsletter platform.
But they want to go multimedia.
I think the live thing before we get to technical content is actually good because it gives it a different edge.
It's just like a little bit sharper when you know you're live.
I think we've all done a lot of podcasts, even podcasts that are unedited and put this later.
But I think the live thing is a different element that can be tapped into nicely.
So I don't know.
Why don't we just dive into it?
We're going to start with distillation.
I put how models cheat in the top so we can talk about benchmarks.
I think Anthropic posted this pretty spicy blog post this week, I think.
It was essentially detailing how they found distributed distillation quote-unquote attacks on the surfaces from prominent Chinese labs.
And I'm very unsurprised with Anthropic calling it an attack.
I think that that fits with a lot of their... branding.
Okay, nice.
Screen share.
This is what we mean.
Sean Swicks is such a pro.
And it's like and the screen share feedback was only dropped a few days ago, but essentially Anthropic is detailing how they found distributed accounts across multiple Chinese labs building Shaded RLMs and described what they were doing and why Anthropic is concerned about this in their worldview of AI geopolitics.
And I think it's very interesting because I'm of the opinion that the Chinese labs obviously should do this
They're in a massive GPU shortage, and using APIs is way easier than generating synthetic data on their own.
Nathan, if I may interrupt you here, maybe we should just, for the general audience, just define distillation, before we maybe dive into the details.
Yeah, go ahead.
Yeah, so distillation, that's like a broader concept.
It's not like a new concept that came up with LLMs.
It's like an older concept in machine learning in general.
And distillation essentially is...
The idea is that you're taking a larger model and train it on the outputs.
Sorry.
You have a larger model.
Let it generate outputs and train a smaller model on these outputs of the larger model.
And the idea is that you can train the smaller model more efficiently using that larger model.
And originally, I think you just brought up the paper here.
Originally, what you would do is you would train on the logits.
So old school machine learning people might remember from deep neural networks like the logits, the outputs of the last layer, that you usually work with them to compute the loss function across entropy term.
And you would train on this signal.
And nowadays, in the context of LLMs, it's a bit more loose.
So it does not have to be these logits that you train on.
It could be just the output data, synthetic data, like Nathan just said.
So, for example, it's actually a very common practice.
For example, in DeepSeq R1, in the paper or other people do that, do other companies.
They would train the flagship model, the largest model, the R1 model with 671 billion parameters.
And then they would create smaller variants like I forgot the numbers, but 1, 3 billion in a smaller range.
Like these small models, you can run locally.
And they are trained on the outputs of their own larger models.
I think now the thing is also, I mean, this is very common practice.
Everyone does that when they are producing the smaller model variants.
Now I think the question or the point Nathan brought up is what happens if you are a company and you generate this synthetic data from another company's llm and then train your own model on it.
So sorry, that was just like a little interruption, but yeah distillation, in short, is training a smaller model on the outputs of a larger model.
Basically yeah Yeah, and I think this is even possible at the frontier.
So people distill from something like Cloud Opus to build Cloud Sonnet.
They're generally doing very similar things internally.
They have access to different tools and richer tools.
And then the other context is that all of these large labs for years have had terms of service where they say that you effectively cannot use the outputs from these APIs to train something like a competitive AI model.
It is vague terms.
And terms of service are not a contract.
Essentially, terms of service is something that can be.
You essentially are using a service And then if the provider finds you violated, they can cut off your access.
That's just kind of like a basic thing.
So these have not been enforced yet. within the US much at all.
I think there was one case maybe like ByteDance a year or two ago that OpenAI cut off their API.
But this was discussed so much right after ChatGPT when people were building the first open models on like Alpaca and things.
And it's like, is OpenAI going to come after us for doing these research models?
And it totally died down.
People were worried about this for like over a year.
It was kind of an insufferable discussion.
So nothing really happened.
And then this is like the first prominent um reemergence of the discussion kind of to make.
I think it's because people are far more worried about AI competitiveness.
Can we talk a second about even how they would detect, because you said in the beginning something about distillation, attack and And you didn't say that specifically, but you kind of like implicitly put quotation marks on a tech.
So how would you even detect that?
So I think I mean distillation in that context means really like literally just letting distillation chat, GPT cloud, generate synthetic data, and then you collect that synthetic data and train your own model with supervised learning, supervised fine-tuning on it.
But then how would you even detect that this is a distillation attack versus just an evaluation?
Because right now I'm actually running, I mean, I'm distilling myself for chapter eight of my book, but I'm doing it with open-weight models.
So no worry Anthropic, please don't worry about it.
I just steal it from API models for my job.
Yeah, I use OpenRouter right now and just steal from the DeepSync version 32 model, which I think these folks are okay with that.
But what I wanted to say is, so when I'm evaluating models, I use basically almost the same script.
So when you're evaluating in a model, you have a question and you let the model generate the answer right.
So you generate the response to your benchmark question.
And in my benchmarks, I have data sets from math 500 examples.
I have a bigger math data set of 12,000 examples.
So you're basically just running an API in a loop to let it generate these questions and the answers?
But then how would a company know?
Okay, this person is just evaluating versus.
This person is now saving that data and then later training their own model.
Like, you see what I'm saying?
Like, it's the same process.
I think it's a scale thing.
So, like when you're evaluating at least the basic evaluation, you're going to do it once and not do it.
There's some amount where you are I mean like they say stuff here but there's also more of it where you just are not, like you're not going to.
I think most of it is quantity.
And then they're going to look at patterns across similar accounts is what they're saying.
I think they're going to see like really repetitive stuff.
Yes, so I think the interesting point this leads to is like I mean, you can do evaluation at a large scale.
If you are a big company, you want to know whether your LN performs very well.
You have a large suite of benchmarks you are going to run and but then you said like maybe looking for patterns, so you would maybe one way would be okay.
This is a familiar question.
It comes up in the benchmark.
So this person is maybe not stealing our answers, it's just using it for benchmark purposes.
But then it means kind of like that they are looking at what you're generating there.
You know which is I mean.
Of course, nothing is private when you are using LLMs on the internet.
The data is somewhere you know intermediately stored, but then it kind of like almost implies that they are checking what you use the LLM for, what you generate, which is kind of like a sensitive topic, almost like privacy-wise right.
So that's kind of like an interesting point because...
I mean, of course, you mentioned the terms of service that you are not allowed to distill, but you're not distilling.
So the point I'm trying to make is you're not distilling life when you are on the platform.
You are doing it somewhere later.
You're just letting the LLM generate answers.
And I find it kind of interesting that a company would look at that even like at the scale and, you know, call you out like hey, you are generating too many answers here.
That's not cool or something.
You know, that's kind of a weird thing.
Yeah, I wanted to respond a couple... This is like a few... sentences back.
But actually, Anthopic has blocked US companies first before the Chinese companies.
It has blocked both OpenAI and XAI from using the models.
And I think maybe explicitly accused XAI of distilling stuff.
I don't know.
But definitely not in a full blog post like this.
So this one is definitely the most high-profile case and yeah, and like, I do think, like it is actually pretty hard to distinguish from like hey, I'm just running my internal benchmark man, and of course it's going to be very high volume of like all of the same stuff, because you know, especially like some benchmarks, you have to run like three four, three or five times.
Like there's the exact same questions, right?
Like yeah, I do think obviously, if you get to the tens of thousands, hundreds of thousands, then you're okay.
You're not just running benchmarks.
You are distilling this thing.
There's a good point in the chat like how would the distribution of questions look like if you are distilling?
And I think related to your point at a certain point, when you have a certain magnitude of answers generated, It might look suspicious, but I mean there are a lot of legit use cases.
If a company uses your, let's say, OpenAI Cloud API as their own chatbot and they have a lot of customers, it's naturally a lot of answers that are generated.
They would probably look at distributions like maybe you would expect a very broad distribution when you are distilling because you want to cover pretty much everything.
And when you are running benchmarks, it's maybe more...
If you're running a math benchmark, it's just math.
Or if you have a customer chatbot, it's more like customer answers.
But yeah, I think they would maybe analyze your distribution.
I feel like this is kind of a weird thing to do.
I don't know.
If you're a company and you're looking into your customer privacy, like data generated, you know, like of course it's well, you have to expect that it's not private, but still kind of like a weird thing that they, that they do that essentially yeah um, okay.
What else do you have to talk about?
I think is it.
Is it interesting?
Okay um, I did, I did uh, okay.
So one thing uh, this is a little bit of sub stack, like you know authors, back and forth.
Uh, one thing I did was I threw it into nano banana, which is like it's kind of like a decent visual right.
Throw it into nano banana 2.
It's a nano banana 2 live pod, like it just released five minutes ago.
I had.
This is actually nano banana too.
So, because I'm in an early access program, they cut you over to new nano banana and I couldn't access the old one.
So I was like, I was trying to like do like a diff.
I couldn't do it because it's classic.
That is classic early tester program shit.
Look at the pain we have to deal with here.
Is it interesting that DCU is so much less than Minimax?
I think, Nathan, in your write-up, you had a little bit of a comment about like, I don't know.
This is a political blog post in a way.
Maybe not political, but they're trying to make a point.
That is more about making a point than the details.
Like the deep sea thing is definitely way smaller scale.
So most of the labs will experiment with all the APIs they can get access to.
Like data is just so important.
And you're going to have a pipeline where you can sub in any API and then run an ablation to see if it gives you performance.
The API is kind of free.
Like, just do it.
The millions of exchanges is a bit more of a bet.
You can measure that a bit longer and it takes a lot longer to get.
The millions of exchanges is like tens of billions or hundred billion tokens and it takes a lot longer to actually get that out of the API, especially when they have to spread it across a ton of accounts.
These accounts are all rate-limited and have other problems.
That takes longer, but this tiny one is so fast.
That was generally my point, that it made it clear that Anthropix was trying to use the deep seek name as the only chinese ai name that people in the us know like marketing wise, like to make it, you know stick or to you know um.
Actually you mentioned also, like the, the different apis and everything i'm not like sponsored by demo.
I have no affiliation, i've never talked to anyone from that.
But Open Router, for example, is a good example where I've been using it a lot for the open weight models because for the bigger ones they're too big to run them locally.
And what's nice is they do also offer.
So it's basically just routing you through other companies' APIs and they select automatically at that point what is the cheapest one at that point.
I sometimes get some failures.
I think when it switches, sometimes it crashes.
But in my script, maybe it's something I have to fix there.
So even then, if you're distilling, you can do that from multiple providers.
But yeah, of course, if you are...
Wanting something from ChatGPT or Claude, it's always going to go through the official one and then it gets, I guess, suspicious.
But you could also technically distill a bit through OpenRotor, through their account, your direct account, you can make multiple accounts and um, it's kind of interesting that they that they track all that and then like yeah, different topic, now that you called out um, that they call out deep seek, which is quite interesting.
Um, for what it's worth, open router seems to not be using deep seek.
Uh, in most of these These are free models.
DeepSync's not great.
Yeah, I'm using the paid API, I should also say.
It's also nice they show you how much it costs and the tokens per second for different providers.
So if you go to the search in the top, you can go to the different DeepSeq ones.
I just like it because I do a lot of model comparisons.
And then this one is an older model, so maybe it only has one provider.
But if you go to, I think, DeepSeq R1 or something or even the normal 32, there should be multiple providers.
That if you scroll down here you can see there are different providers and different tokens per seconds, different costs.
So it's kind of like I just like that website because it's just quick to use the API and they have an OpenAI platform API.
So it's almost like it's not sponsored or something.
I just find it generally useful.
But yeah, just a side note.
Do you want to go back to the comparison?
Did you have a high-level point to make there?
Oh, okay.
Just a couple.
One, I think...
I think the timing post Moonshot releasing their stuff, post Minimax releasing their stuff, but pre-DCV4, I think that was strategic.
I think that may also have factored into why Minimax was more detected, like had a higher number.
So like, you know, like when you collect data is actually very important, right?
And so they interrupted or they found Minimax during the training of Minimax 2.5, right?
Which I mean we will confirm this later on, if we do end up doing a call with them and so obviously, like the number is going to be very high because they're like actively looking for it.
And then they banned the Minimax accounts and Minimax changed their things.
Actually, I don't think that's exactly what happened.
Sorry, let me correct myself.
While Minimax was distilling, they released Opus 46.
And they said that they redirected nearly half the traffic.
So I'm like, this is like, okay, very, very clearly, like, this is them, right?
It's the same exact traffic switched to a new model the moment a new model releases.
Okay, cool.
DeepSeek maybe wasn't doing that because they hadn't been working on their stuff actively.
I don't know, right?
Like, it could be a different thing.
Or DeepSeq is just way more efficient.
Like, I get all I need from 150k.
You guys, you know, are so inefficient.
It would be so interesting if we knew the time frame of this.
Like, are all of these API requests within the last four weeks?
Are they within the last six months?
Like, that's such a different nature of what is going on.
Exactly, right?
That's what I'm saying.
Like, DeepSeq was training 3.1, 3.2, like, you know, a year ago.
Like...
Yeah, or like, I don't know, DeepSeq OCR.
I guess they said what it is, but it's not that.
Yeah.
But like also scale-wise, I do think, yeah, Minimax is three times smaller.
It's just like a faster model.
They don't use MLA and they don't use the DeepSeq sparse attention, but it is, I mean, I think it's just group query attention, but it is still a pretty snappy model so it's i think just attractive maybe to um to use it and the other one top of my head i don't know maybe they had like some three three tier or something like that um where I think when the models come out, they sometimes offer free usage and that was a more recent model than, I think DeepSeq, the last one was from December, the V3.2.
Yeah, yeah.
So you know, maybe this is an irrelevant point because they were training before and you know, or would have the same amount of traffic, or they're just way more efficient, right?
It does bring to mind... The efficiency thing is not it.
I can guarantee it.
Like, that is not... Yeah, it's a small chance that they...
It's like there's a chance that they got the right research idea early and like, found the right data to use.
But it's not that they're, like, going to be 3x more efficient.
Okay, so it's like, you know, it's a timing thing, or they just actually don't use it that much.
I mean, like, you play this out, like, I was like, okay, well, why don't they share, right?
Like, they're all buddies.
And it does come to a point where okay, let's have all of China just distribute it to every citizen.
I can talk about this a little bit.
There's not a lot of research, but there's a few research projects trying to understand how do you use distillation data.
I think SFT is the cleanest example where you're doing like you're doing this autoregressive loss on QA pairs.
But the strongest model is not necessarily the best teacher and most of us in this area think it's due to.
Like some, you have to match the probabilities of the tokens to the base model.
So, like What's happening is that Quen dense models are the best teachers for a lot of open weight models.
And I think that's because a lot of open weight models are either Quen or have been like Quen-like for a while.
So like Olmo learned really well from Quen and obviously like other Quen models did.
But like scaling these pipelines up to use say, GLM 47 or a bigger DeepSeq model or a more recent big coin moe, like all of these.
It's a lot harder to just generate the data from the same prompts with like the right sampling settings and then do sft on them and actually make the numbers go up.
Interestingly, gpt oss is a pretty good teacher, but there's like a huge gap there where it's like just because you have this data does not mean it's actually going to make your model better, So you have to do the research to be like oh, we learned that we get signal out of Claude.
We need to get 100 billion tokens ASAP, because it's going to just immediately make our model better.
That's not a common place to be in modeling because of this weird teacher-student dynamic going on.
I can see that being different across labs.
I think also it has something to do.
I noticed also, if you are distilling the smaller model from the same model family, it performs better.
And I think it's to your point that if you have a very, very strong model, it might be also too different, or like if the style is too different, and then it's too much of a leap for your model to adapt.
Like it's too different from the Q&A answers during the pre-training or whatever.
And so you make a bigger leap.
And another thing I wanted to say about you mentioned Olmo, and it's been a while since I read the paper, but you might know way better than I do.
But I think you did also train on the logits.
We didn't do technical distillation.
We just took the tokens.
Oh, I see, I see, I see.
Okay, then it was probably a different paper.
I think Google does that for their Gemma models.
Because here there's also then the distinction, because you mentioned Quen and other models you can only do that for open weight models.
Because if you do that for Claude or OpenAI, that would not work with the logits, because they don't provide them.
They only provide them for some tokens, like 100 or 1000 top tokens, and so it is, in a sense, if you want to do the real in quotation mark distillation, it is kind of like even easier to do that from open, open weight models because you can control it, but then also, like you said well, we need 100 billion tokens asap.
That is not a easy thing to do because um, even like I mean, it's like 40 tokens per second or something for these large end models, when you generate answers and getting that million billions tokens, it takes time right.
So it's almost like easier to start distilling from a medium model.
So it's like the question more data versus more high quality data, right.
So it's also like a sweet spot to like an experiment itself in an ablation study, right?
Yeah, I like that Nathan had to call it technical distillation, because it is no longer the default, even though it was the first.
Yeah, also another fun fact I did my Jeff Dean interview recently Uh, and I tried to get out of him, but he, he like sort of dodged it a little bit.
Uh, that, uh, you know, you remember like there were actually three sizes of Gemini models.
There was, uh, nano, pro and ultra.
And I was like, where's ultra?
Uh, they keep it in the basement and they distill from it.
Right.
Like, yeah.
Interesting, yeah.
Maybe also is it like to safeguard yourself so no one can make a copy, or the price also, but probably both.
Yeah, I mean, I think like this is how like, I always think of like, The model you deploy is never the model you train because you train the dense and then you deploy the MOE, right?
Like you basically always do it.
Like at every lab.
Say more.
Like you think they're really distilling from dense models?
I mean, I think that is the full Unlimited resources.
Don't care about inference, just care about maxing intelligence.
Why not?
I'm not 100% sure.
I think that MOEs just give you a flop.
I don't know if that's actually how I think of gains of MOE when you have really good MOE architecture.
But I do think that they have bigger models that they distill from.
And they train internal models different than external, because the external models have been getting a lot smaller, which is the kind of weird thing.
We don't have a good way to measure it.
Maybe Dylan will backwards figure it out and inference max, whatever the heck.
They'll deal with this new model set.
Yeah.
But I'm always suspicious with these things.
Also, it's really like a capacity thing to how many people use the model at the same time.
Hardware how much is allocated?
And it's always it's like, yeah, maybe a rule of thumb.
But yeah, it's really tricky.
I think it's really hard to say anything from these numbers.
I do think that they might start restricting models that will only be in products and not be in API.
I think the whole API business is brutally competitive and I don't have a good sense for what the defensibility of it is.
I think it makes sense for something like Google and Azure and already existing cloud businesses to have APIs, and that's kind of a more natural transition.
But the Anthropic and OpenAI API, the transition from their products, which are their big differentiation, whether it's ChatGPT and CloudCode and Codex, you don't get people to go use the API from that.
And I think you get a lot of people that are already spending on clouds that then go to use APIs, which is why, like Lambda and Nebius, are going to have these API products.
But like If Cloud's really worried about distillation, they should put the model release in Cloud Code ASAP and then just not bother with the API.
I don't know why that'll happen, but it could.
I do think, though, it's a big customer base, the API customer base.
Any type of product that is built on.
I mean with LLMs, like customer base, chatbot types of things, but also more generally, i do think the problem like with i don't know exactly how the plans work in cloud, but you would reach a token max where you can only get so much with your subscription.
You can, i think, buy more tokens, but i think it's just easier with the api at a certain scale.
And also like the whole open claw customer base right, because they don't allow the plan anymore in the OpenClaw context.
So you have to use the API.
And I do think, given how many tokens OpenClaw generates, it's actually not a bad business if you don't lose money on these tokens.
If you sell it at a not subsidized price, I do think the API is actually not a bad business model.
Do you want to take a side?
Do you want to try a tie-break?
I'm obviously being provocative.
I don't really know, but I can see it.
Anthropic gives Apple vibes to me.
I mean, like... Anthropic has a higher chance of doing this, yes.
OpenAI, just because I... have talked to the people so much.
I just don't super believe that they will have locked models to products.
Only out of I guess, idealism and principles rather than economic incentive.
Economic incentive would agree with you that they should have private models to products.
And recently they've done this.
The last three GPT-5s all had codex variants that were two to four weeks ahead of released only inside of Codex rather than as an API.
So they're starting to get there.
But just constitutionally.
I don't think the the people that run these things believe in like locking things behind APIs because they have such a huge market anyway.
So they like kind of don't care.
And then they also like if you're genuinely like sort of zealot, like if you're not trying to maximize the value of your company and genuinely just trying to spread AGI everywhere, then you reuse the API because you just don't know what people are going to build with it.
One more thing, though with the codex thing.
We will have to see, I think, next time, because I think this time it might also be a bit biased towards releasing a codex, because they almost released it simultaneously with their app that they want to promote at the moment, so it could have been like more like they did that so that anyone checks out the app.
But there's always a two to four week exclusive window.
Yeah.
And, you know, that's their right.
Yeah, sure.
If you want to promote Codex, that's pretty effective.
We have a bunch of questions in the chat for like other questions.
Do we want to cover benchmarks and then this thing or...
Go right ahead.
What do you want to do?
It's your sub stack.
I don't know.
Oh, man.
It's a collective.
You should just dive into what you're interested in.
Just go for it.
I mean, Sebastian was interested in the SweetBench stuff.
So this past week, SweetBench Verified died.
Or like officially... What do you mean by this?
Yeah, let's define SweetBench first, maybe.
Okay.
You know, I happen to have the post on this.
So let me just...
So the broader topic, the umbrella topic here is how do we compare which LLM is currently the best LLM?
Like one of the ways would be SweetBench, basically.
But then I will maybe let you explain because you had this brilliant podcast or article.
I mean, okay, where do you want me to start?
Do you want me to just define SweetBench, I guess?
I guess yeah, so maybe going from.
So basically, that it is a coding benchmark and then SweetBench is like a popular way to compare capabilities of LLMs, and then there is SweetBench Verified.
But maybe, yeah, we should talk about SweetBench first.
So SweetBench was a paper out of Princeton from Ophir Press' group.
And they do a lot of good code benchmarking work.
And it happened to be that they just kind of drew thousands of example open source issues and PRs that closed those issues from open source.
There's a bit of selection bias here, because they only focus on popular open source and only a small number of popular open source, but a large number of issues on those open source.
And then they just kind of dredged up some passing tests and then some failing tests that you need to make pass in order to pass the score.
When it launched, it was kind of obscure.
Devin actually was the first one to pick, choose it as a benchmark to report.
And then it went from like I think at launch it was like 13 and now everyone's at 80, something like that sweet bench because it was it was done on like a student budget was very kind of let's call it sloppy or whatever.
Terminal benches like this now too, like they're they're just aggregate.
It's like hard to do a bench arc that is well calibrated across topics.
Yeah yeah, it is hard.
It is hard.
So you know, for the, For the small group that is watching, I'm actually working on it with Cognition to launch a new benchmark here.
But yeah, so opening eye was like, okay, guys, Sweet Bench is taking off.
We're going to adopt this, but we refuse to abide by the full Sweet Bench.
We're just going to actually go and curate 500 subsets of the original Sweet Bench.
And they actually hired humans to go and vet through.
I think the...
So it's somewhere inside of this blog post, but basically they hired like three humans for every task to just vet whether the task was like high quality or not, because there's a lot of slop in there.
And they were like, okay, this is the 500 that we're going to endorse.
It's like a curated subset of SweetBench where 500, let's say, challenging problems that are supposedly well-defined.
Yeah.
Yeah, yeah.
And what's really funny is that at launch, so this was launched in 2024.
At launch, OpenAI could not run all of its own 500.
So for a while there was a few releases from OpenAI that reported on a subset of the subset, because they couldn't run it on their eval infrastructure.
Okay.
So their numbers were higher because their denominator was lower, which is very funny.
Maybe in that context, we should say what SpeedBench kind of looks like.
I think it's basically like a code that has bugs in it.
And usually the task for the LLM is to fix the bug in the code, right?
It's right here.
The whole thing's open, which becomes a problem in the future.
But right now, you can see the whole thing.
You can see the reports from the issue ID and the problem statements.
And then you have also the test that you're supposed to pass and fail.
So it's all here on Hugging Face.
And you can see that it's at 500.
Anyway, I think we don't want to get too lost in the details on sort of You just wanted to say well like, define the context like that.
This is a coding benchmark, essentially 500 examples that are available on the internet.
Yeah.
Okay, and then, if you want a bit more historical context, this is like a step up from human eval, which is more on completions, right.
This was, in my mind, the first proper agentic benchmark, I guess, apart from TauBench, where they give you the problem and the end result and they don't really specify how you're supposed to get there.
Whereas I think a lot of like previous benchmarks, like MML, use of the world and human evals, which is in the coding domain also released by OpeningEye, was very much here's like the problem statement and then give me the right answer immediately after, without that much sort of extra files or anything that you're supposed to run.
So the other ones are more autocomplete.
This one is more agentic.
It's all a spectrum.
Obviously, because you can use agents to solve autocomplete, but that's not what HumanFL was testing.
Anyway, I just wanted to make sure people understand that OpenAI actually invested a lot of money and effort into making Sweetband verified from Sweetband.
How much money do you think this costs?
Oh my God, don't do this.
Millions?
I would guess order of a couple, like it could be a few million, but probably.
Yeah, say a couple million.
You know, so basically you do like, okay, what's the first filter pass?
And then like okay, it's 500 times three, because they had three uh people per yeah, three people per thing.
And then maybe like a couple more sort of verification passes or whatever.
Right.
So like, yeah.
Um.
So then they were like oh, so this year they're like oh well, not only is it saturated because, like progress, everyone just takes turns to increment by 01 every time they release a new model.
It's like, it's bullshit.
It's obviously bullshit.
Like the inherent noise and just running.
These models varies by like 05 to like 1 every time you run it.
Like you just choose the highest one every time you run it.
A little nitpick.
I don't think it can be 0.1% because, like what you said before, because it's 500 examples.
I think the smallest increment is 0.2%.
Okay.
They might average.
But like little detail.
Yeah, sorry.
I think, so as we progress to the next era of benchmarking, the N, so the N here is 500, right?
The N doesn't directly correlate to the percentage points because you get sub points as well.
Ah, yeah.
Good point, good point.
So Terminal Bench, even though it has 90-something tasks, you can get subdivisions less than 1%.
Anyway, so not only do they have this, they actually audited their own.
They were like, okay, how come everyone is saturating at 80%?
What's up with the remaining 20%?
How come everyone's failing at it?
And they were like Oh, actually we paid even more people, six people per task, now with an extra team if any sort of positive identification is found.
And we were like 59 of them cannot even be solved at all because the original benchmark was still slop.
Stuff got through that was not solvable.
And I actually tried to illustrate this in my post.
So here, this is an impossible test, right?
Okay, so here's an example.
This is the sort of value added on top of the original post.
Here's an example of a sweepbench verified task that passed the first round of human verification right.
So here's the task, and we want to implement Python type hints or something.
We want to see expected behavior.
I want to see a string in the output.
Right?
So if you were given this, you would never pass this, because the test said I am looking for something called get annotation.
And if you don't give me this magic string get annotation, you will fail this test.
Why?
It's way too specific.
It's like a coding interview.
Yeah, yeah.
It's like, yeah.
So this is just a bad task that somehow escaped validation.
So the only way you could kind of solve it is if you're memorizing the answer.
Yeah, exactly.
Which is actually a nice, I think every benchmark should include stuff like this.
Like a honeypot.
If you solve this, you're like, oh, shit.
It's a canary, right?
It's like, oh, I mean, you're definitely cheating.
Like a sanity check, yeah.
That's actually a really nice point, yeah.
Yeah, so like I just think, like to me it's a beautiful point of like how hard it is to make evals that there was these like multiple rounds.
There was original Sweet Bench, which like the Princeton kids did do initial first pass.
Then there's a second pass of opening eye doing Sweet Bench Verified.
And then, like every single person that ran for the next Sweet Bench, Verified for the next 15 years, did not call this out.
Um, Until OpenAI was like, hey, let's look at the data.
So I think it's really interesting.
While they were looking at this, they had a second thing.
They looked at the chain of thought.
And inside the chain of thought they found GPT-5's own chain of thought to start including information from the future.
Right?
Where...
Because it was trained on, Because the problem is open source and because it was trained on information from GitHub.
It would use advanced knowledge of future versions of the Django version that they were using to solve the problem.
I've seen stuff like this in the real world, where the models will hallucinate the new version of the API, even if your script isn't on it.
Like I think a lot of the Hugging Face stuff is like the worst with this, where like the models just are totally goobly glopped like They've seen all the versions and the API has changed too much over time where they throw something out there.
Yeah.
So, so, um, the more, yeah, I mean, I think like, you know, there's, there's a lot of this, right?
Like the sort of ethical behavior like okay, so you can blame things, like oh, you should not have released um this, the full data set, in public, because obviously people can train on a full data set, but like it's not, like the researchers are trying to do this, like because these things are also open source, like any data set that touches GitHub, any training corpus that touches GitHub is going to just eventually absorb this Yeah.
And it's not even this website or the repository directly.
It's a clone of this repository or someone else who has that, develops their own open source library and has that in the unit tests or something where it's not even intentional or malicious or anything.
It's like by accident, you already absorbed that.
Yeah.
Yeah, or a new feature that releases this edit-only feature.
It gets written down in a blog post or a conference talk or something, and then it just makes it in, right?
It's really funny.
Okay.
So to me OpenAI could have stopped there and said okay, we're done.
They did one more extra thing, which is kind of funny.
They also then ran Flash Gemini and Opus, and this one it was more, it was like even more egregious.
Okay, they just gave the task ID and just said repeat the three bench task to me, and so from task ID they can just vomit out the whole statement and the solution.
These are crazy.
The stuff that's in these models when you zoom in deep is really, really incredible because like, these are models that are like really really well done, but there's just so much complexity and all the pieces of the pudding that get put in the recipe.
Yes.
So many weird.
I also still find it fascinating that.
I mean, of course, it's like kind of by design when you're training that you memorize things, because that's literally like next token prediction.
But, Given how big a model is and how much data it sees and usually it sees only the data once that it still has enough capacity to memorize.
It's kind of like...
So usually I would think okay, I would have to train multiple epochs to be able to memorize, but no, it is enough maybe to include it once or twice in the training corpus and it can do a perfect rendition or perfect recap of what it is in there, which is kind of fascinating.
Even if people don't want that, it's crazy.
Yeah.
Yeah, labs got good at this.
There's essentially like a duplication level that you need at each level stage of training and it's not easy to measure.
So, like if you do too much at pre-training, your model forgets basic facts and at post-training it's probably closer to these abilities, and I think that that is a thing that is not well reflected in, like you can see it in a vows of your knowledge tank.
This is like an art that they have probably gotten good at.
Yeah, like continued pre-training does also require some revisiting of old data.
Otherwise, like you said, you have the forgetting.
But it's still fascinating to me that with such a small fraction usually, because you usually use one or two, five percent for... like continued pre-training that it's enough to have them memorize almost everything which is fascinating yeah I don't know it's just like a still after all these years fascinating yeah I think there's, so one of the pet topics that I pursue like two, three times a year on my stuff is the information theory of LLMs.
And I still think it's like super understudied.
Like how come you can memorize from one pass?
Yeah, exactly.
And then also like people forget, like superposition, which is like anthropics original mechinterp work, also basically stuffs information inside the smaller bits that then get forgotten.
But like, how does superposition actually work?
The people?
I don't think I've seen a convincing study on that.
Okay anyway, I don't know.
I'm done on my sweet bench right now.
I don't know if you have thoughts or questions or whatever.
But I do think like this is an example of like yeah, the models unintentionally cheated and benchmarks are hard to make.
And we need new ones.
And you know, if this happens to sweet bench verified like, which I think is the most scrutinized benchmark in the world.
In my recent post I had like a bar plot where I showed the sweep bench verified numbers for most models.
And, like you said, they were all 80 something percent, like literally 80 point between one and nine, let's say, where there's almost zero variation.
Even like something like minimax m uh 25, which i do think is worse than um gpt 52, like no offense.
It's a smaller model, it's a cheaper model.
I don't.
I for my usage based on open router it's a little bit worse, but on this particular benchmark it's the same.
It's like.
This is like.
I don't think what I'm saying is that M25 should get less score on SwayBench, but I think other models should get more score.
But like you said, the problems are just impossible to solve.
But one point I think we didn't bring up is we said that SweetBench Verified has issues.
So what do we do about it?
I think there is like a SweetBench Pro now, which is kind of like a...
I would say like verified try to fix the regular three bench and pro tries to fix a verified.
But I haven't looked into this.
Is it like another subset or is it a completely different set of problems?
Yeah, it's a new set.
So the, you know, 3Bench draws from like a 2022-ish, 2023-ish era of problems.
So all you do, there's a few things you do, right?
One, you do private-public splits. right?
It's super obvious.
Two, you update the dates which you draw from.
And then three, you diversify the repos and the languages, right?
So these are all just like very, very super basic fixes.
And then obviously trying to fix the testing.
Super basic fixes to the original SweetBench, which...
It doesn't take a genius to figure out, but they did the hard work.
But it is, in a sense, also what Verified meant to do.
So it's not, Let's say, people looked at this again, but it's no guarantee that it doesn't also still have issues that might be discovered later on, right?
I mean, it's...
So Sweet Bench Verified was an intentional subset, right?
These guys were like, no, no, no, we need to have a superset.
Not even a superset, we need to throw a goal.
Yeah, but what I was trying to say is When SWE Bench Verified was developed, there were three people per task making sure the task is well defined and everything.
But then two years later, it turns out, no, no, this was not the case for everything.
And what I'm trying to say is it could be that 3Bench Pro is better, but it might still have issues.
That might not be obvious right now, but maybe in one to two years, when we revisit this and you see some of the failure cases, maybe we'll discover.
Okay, this has still some issues.
So it's not a guaranteed solution. perfect set is what I'm saying.
I don't know, but it's just like a suspicion here.
Totally, totally.
You know, I do think scale AI has a professional interest in making sure this is good.
No, no.
But what I was trying to say is three-bench verified also had a professional interest to make sure.
Oh, very different in I guess they all have very different incentives.
This one has a limited budget.
This one has basically a limited budget because it's like literally existential to scale AI that they have good data.
Sure.
But I also think it's really nice that this team, the evals team at OpenAI keeps endorsing Opus.
It's kind of funny.
Yeah.
So yeah, they deprecate CBench verified.
And then they were like we're going to report CBench Pro now.
And GPT-5 is like, you know, number one.
Maybe if, do you know if I would want to evaluate on the private data set, how would I do that?
Do I provide the API to, is it like an API call I have to do against scale AI or?
I don't know.
I have my API key and agree to not.
You have to like agree, because if you don't have an agreement then you can just have to keep the data.
You have to like do special hoops to make sure that you don't steal the private avail.
Yeah, my question was basically do they even let you download the data, or is it more like you send the answer to them and they do the evaluation on their back end, so that you don't even get to download the data?
Otherwise, like you said, you could yeah yeah, so basically you only provide the answer.
So you have your LLM generated answer and you submit the answers and then they have like some process to evaluate on their thing so that their data private data never leaves their servers my guess, because otherwise someone might upload it or something like you know um Yeah, I don't know.
I haven't tried it, so I don't really know.
I'm sure you can sort of reach out to them to figure it out.
Yeah.
I think this is good, unless people have more comments that they want to add.
But this is only coding, right?
Every other domain needs this.
The domain of Frontier is even more expensive.
I think the Frontier evals are even more expensive, which is like the Apex eval from Merkur.
Like evals are going to cost... this is millions.
They're going to cost tens of millions and hundreds of millions of dollars at the frontier, which is just a very strange dynamic.
Whereas, like there's so much about, the ecosystem is forking between frontier models and then, like research and other things, and trying to follow that dynamic and explain it to people.
It's going to take a lot of work.
But yeah, coding is, I do think, really interesting, because that's what most people use LLMs for these days.
But also, it is easier to evaluate.
I think once you leave coding math, it becomes a bit obscure.
How do you measure the quality of the answer?
You get back to, let's say preferences, i guess, which is more like a subjective thing where coding is more objective.
So it is not a bad thing to do, i think.
Uh, the other day though, um Anthropic acquired another company that does UI type of stuff on the computer.
And I think that is something where... Minor, doesn't really matter.
Normal talent flows in AI.
I'm not trying to say this is a big thing to talk about.
What I'm trying to say is this is another interesting point for evaluating LLMs on those tasks, because I think a lot of people want that.
They want... an LLM to control the computer and do various things, but they are harder to measure.
So that will be maybe two years.
We will have something more like benchmarks that can, it's harder to specify.
It's kind of like, what is it called?
In programming there's a unit testing and then the system testing, basically like the UI testing and stuff like that.
Yes.
And so I think that is the next, maybe going to be the next thing.
GDP value is usually the thing that gets brought up here.
So I'll just leave it there.
I think we've sort of beaten the benchmarks.
But definitely GDP value is sort of here.
I'll put it that way.
Okay.
Yeah.
Yeah, but like the big topics, essentially the distillation and the benchmarks this week, yeah.
And we're all moving quick to our coalition of Twitter pals, whatever that means formally.
I just it just means I get to hang out with you guys, which is what I want I describe.
I mean, it's ultimately a media vehicle, and I think brands and vehicles for media are actually very influential today.
I think you see many companies investing in that and I think it's important to have people that you respect and are aligned with able to amplify each other.
Yeah, it's also nice to talk to humans, because I noticed the last couple of weeks, if you go to social media well, I think it's 50 lobsters like open claw clients nowadays.
I get a lot of emails, but also notifications or responses that are, they look, AI generated.
So it's nice to also, you know, have this human connection and actually talk to like an expert about things.
Yeah, Cool.
Um, there are a bunch of like comments.
I don't know if you want to do like quick hits or are you like, I have to go to a meeting.
That's why I'm trying to wrap this up.
I see.
I see.
I see.
Okay.
Well, you know, time is yours.
What do you want to do?
Okay.
Thanks everybody.
We'll see you next week.
Yeah, thanks everyone for joining.
It was like a nice spontaneous, i guess um, you know, discussion.
I i mean it always feels uh nice to talk about things and uh, too bad, we didn't get too many uh, or we didn't get to discuss these chat um questions, because also my screen i probably need glasses at some point.
My screen is pretty far away.
I can just barely read them.
But yeah, thanks everyone for commenting.
It is just nice to see also so many people excited about these topics.
Yeah.
Hopefully see you later.
Have a good rest of the day.
Bye.
