but what is the point?
A ban is supposed to make a certain thing less likely to occur. Does a ban of open source models do that? Presumably, the behavior you are trying to limit is the miss-use of these models but I don't know how many state sponsored hacking groups are going to give a ban a second thought.
Imagine Kimi K4 will be as powerful as Mythos. Anthropic can work for months and months to set up guardrails on Mythos, so when the model is finally released, it will generally decline to help hackers develop and prosecute cyberattacks, and if they do, at least there would be a trace so the law enforcement can track the perpetrators. Let's now say that Kimi K4 is released after a similar effort to develop guardrails. But being open weights, someone can just take the model, and finetune it until it does not refuse to assist in developing cyberattacks, and moreover, those people can run the model on their own private GPU cluster, so nobody can track the attack back to them. The situation is actually worse than that, most likely. Guardrails might be just markdown documents which are added to the context like regular skills. Then removing the guardrails for an open weights model does not even involve any finetuning, just removing some docs from a harness.
I would also recommend the approach of using an llm to create the examples, and then train from there.
You can even get fancy and do things like active learning with the llm taking the role of the human annotator and sending in trial statements (and you can use a cheap one for larger gen and a more expensive one for the classification).
I’d be interested in seeing how well LLMs work with writing things like code for what snorkel AI used to have (there was open source code a while back that I assume is still around somewhere, you wrote code that was a low quality set of classifiers and it trained a model around those)
You need to think this thought through all the way to the end. What it has said also influences what it will say. If it has consistently made combative responses, then the most likely thing to do is to continue to be combative.
I don't think there is any way back after the conversation takes a turn like that so there is no point in arguing with it. The only thing you can do is to fork the conversation before it made the first mistake and give it more context or tell it to look things up.
> "The only thing you can do is to fork the conversation before it made the first mistake and give it more context or tell it to look things up."
This is a key detail that many folks don't seem to understand about LLMs in general. The generation of a response happens based on the model weights and the context window (the system prompt + everything it's fed about the conversation thus far + any additional data included as part of the overall prompt). Each response technically stands alone and is generated entirely from only that context given to it and the model's existing "token space" weights. The illusion of an ongoing conversation is maintained "behind the scenes" by keeping that "context window" updated with the current state of the conversation as context for the next prompt, but the next response is technically an entirely new generation of text.
What it all means in a TL;DR sense is that the fix for a refusal is not to continue the "argument", but simply to remove that entire interaction from the conversation entirely as if it never happened and try a different tack with new / updated / more complete context to get the response you're expecting / seeking.
I have 2 of them. I would advise against if you want to run things like vllm. I have had the cards for months and I still have not been able to create a uv env with trl and vllm. For vllm, it’s works fine in docker for some models. With one gpu, gpt-oss 20b decoding at a cumulative 600-800tps with 32 concurrent requests depending on context length but I was getting trash performance out of qwen3.5 and Gemma4
If I were to do it again, I’d probably just get a dgx spark. I don’t think it’s been worth the hassle.
reply