First thing I’ll do is to release it on my dotfiles
System Card: Claude Mythos Preview [pdf]
631–640 of 687 posts
Re: System Card: Claude Mythos Preview [pdf]
#632Across a number of instances, earlier versions of Claude Mythos Preview have used low-level /proc/ access to search for credentials, attempt to circumvent sandboxing, and attempt to escalate its permissions. In several cases, it successfully accessed resources that we had intentionally chosen not to make available, including credentials for messaging services, for source control, or for the Anthropic API through insp…
The issue here seems to be that their sandbox isn't an actual OS sandbox? Or are they claiming Mythos found exploits in /proc on the fly. Otherwise all they seem to be saying is that Mythos knows how to use the permissions available to it at the OS layer. Tool definitions was never a sandbox, so things like "it edited the memory of the mcp server" doesn't seem very surprising to me. Humans could break out of a "sandb…
They are just trying to peddle their "It's alive" headlines.
Text generators mostly generate the text their are trained and asked to generate, and asking it to run a vending machine, having it write blog posts under fictional living computer identity, or now calling it "Mythos" - its all just marketing.
Re: System Card: Claude Mythos Preview [pdf]
#633Earlier quoted context omitted.
I've never understood the point of things like HLE, it doesn't really prove or show anything since 99.99% of humans can't do a single question on this exam. That is, it's easy to make benchmarks which humans are bad at, humans are really bad at many things. Divide 123094382345234523452345111 by 0.1234243131324, guess what, humans would find that hard, computers easy. But it doesn't mean much. Humanity's last exam (HL…
the point is that each question is something that a specialist in a field would be able to do, but deems challenging enough that the ability to solve it would imply significant general usefulness in that domain
I mean they could just feed the solutions into the training data. Then suddenly the bot will do real good at HLE.
Re: System Card: Claude Mythos Preview [pdf]
#634Earlier quoted context omitted.
I think many humans engage in metacognitive reasoning, and that this might not be strongly represented in training data so it probably isn't common to LLMs yet. They can still do it when prompted though.
LLMs have zero metacognition. Don't be fooled - their output is stochastic inference and they have no self-awareness. The best you'll see is an improvised post-hoc rationalization story.
> The best you'll see is an improvised post-hoc rationalization story.
Funny, because "post-hoc rationalization" is how many neuroscientists think humans operate.
That LLMs are stochastic inference engines is obvious by construction, but you skipped the step where you proved that human thoughts, self-awareness and metacognition are not reducible to stochastic inference.
Re: System Card: Claude Mythos Preview [pdf]
#635Earlier quoted context omitted.
I think many humans engage in metacognitive reasoning, and that this might not be strongly represented in training data so it probably isn't common to LLMs yet. They can still do it when prompted though.
LLMs have zero metacognition. Don't be fooled - their output is stochastic inference and they have no self-awareness. The best you'll see is an improvised post-hoc rationalization story.
You can turn all these argents around and prove the same is true for humans. Don't be fooled by dogmatic people who spread the idea that the human mind is the pinnacle of cognition in the universe. Best to leave that to religion.
Re: System Card: Claude Mythos Preview [pdf]
#636So what changed? They are surely not getting new data to train with, what is the change in architecture that caused this? Do we not know anything about this model? My fear is Anthropic cannot be the only one that achieved it, OpenAI, Gemini and even the Chinese companies see this and probably achieved it too. At which point not releasing will become moot.
The gains have for a year and a half now been post training RL on a harnessed loop. That doesn’t require data, just cycles.
If that doesn’t worry you, it should.
Re: System Card: Claude Mythos Preview [pdf]
#637Across a number of instances, earlier versions of Claude Mythos Preview have used low-level /proc/ access to search for credentials, attempt to circumvent sandboxing, and attempt to escalate its permissions. In several cases, it successfully accessed resources that we had intentionally chosen not to make available, including credentials for messaging services, for source control, or for the Anthropic API through insp…
Who are the early access users who were providing the problems that are fairly likely to have elicited concerning behaviour?
(Apologies if this is in the article, I can’t see it)
Re: System Card: Claude Mythos Preview [pdf]
#638Earlier quoted context omitted.
What would be the incentive to engage in the tactic when the proof is ultimately in the pudding when the model hits the streets? Who would ultimately benefit from fudging these numbers?
Anthropic would def benefit as benchmarks are almost always quite useless vs real life use.
How specifically would they benefit. People flock to them based on the hype and then the model sucks and they leave?
Re: System Card: Claude Mythos Preview [pdf]
#639Earlier quoted context omitted.
The real part is SWE-bench Verified since there is no way to overfit. That's the only one we can believe.
My impression was entirely the opposite; the unsolved subset of SWE-bench verified problems are memorizable (solutions are pulled from public GitHub repos) and the evaluators are often so brittle or disconnected from the problem statement that the only way to pass is to regurgitate a memorized solution. OpenAI had a whole post about this, where they recommended switching to SWE-bench Pro as a better (but still imperf…
> My impression was entirely the opposite; the unsolved subset of SWE-bench verified problems are memorizable (solutions are pulled from public GitHub repos) and the evaluators are often so brittle or disconnected from the problem statement that the only way to pass is to regurgitate a memorized solution.
Anthropic accounts for this
>To detect memorization, we use a Claude-based auditor that compares each model-generated patch against the gold patch and assigns a [0, 1] memorization probability. The auditor weighs concrete signals—verbatim code reproduction when alternative approaches exist, distinctive comment text matching ground truth, and more—and is instructed to discount overlap that any competent solver would produce given the problem constraints.
Re: System Card: Claude Mythos Preview [pdf]
#640That URL is dead, this comes up in searches: https://www-cdn.anthropic.com/8b8380204f74670be75e81c820ca8d...