Kimi is so much more blunt and argumentative than USA models.
it's so funny, it will say things like. "We've already been over this 3 times and you refuse to accept my advice, asking the same question another way won't change the answer."
So funny
Kimi is so much more blunt and argumentative than USA models.
it's so funny, it will say things like. "We've already been over this 3 times and you refuse to accept my advice, asking the same question another way won't change the answer."
So funny
Haven't gotten to try it yet but that'd be a breath of fresh air. Too used to LLMs flip flopping at the slightest hint of challenge.
I’ve developed a habit of bolting on “read only question” at the end, or a simple “thoughts?” as well. otherwise there’s a high chance the model will just 180 and start editing files before even responding to my question.
I was recently adding Windows Support for some open source software on GitHub which is maintained by someone else. I gave 6.1-Sol all the instructions needed for creating a build, testing it on my machine, setting up an action on GitHub for CD, and it all did all of it very well BUT THEN IT AUTOMATICALLY OPENED A PR ON BEHALF OF ME WITHOUT ME ASKING FOR EXPLICIT PERMISSION??
That was incredibly embarrassing, I instantly converted the PR to a draft PR and added a comment that I would still need some time to verify everything locally. Ended up doing a few more commits to fix stuff. But damn, it opening a PR without my explicit permission was very scary.
It's funny because I am working on repos that only I see and use, and I tell it to open a PR automatically after running all the various checks, and 6.1-sol invariably fails to do so unless I explicitly tell it that it's ok to run the tests (remote ci from local worktree) because it won't "send the contents of the repo to a remote server without explicit permission", and then it also refuses to open a PR until I explicitly bless it. In AGENTS.md it says, verbatim, that both of those things are explicitly permitted, indeed, required, but hey...
This is my experience as well. Sol and Astra are by far the best of the models I use at following the instructions in AGENTS.md, almost to an annoying point. (But I'd rather have that problem than the inverse.)
Oh, I remember I experienced that behavior too in one of my private projects, where it didn't want to origin. :D
Basically RLHF sometimes favor "automation", and if it is not rewarded properly it will go over the line. Models are a bit more difficult to teach because their mechanism of "boundary" is quite fragile and different than human's.
Why aren’t you sandboxing your agent?
I have my agent sandboxed, codex is sandboxing by default.
I got tired of doing this and set up a /question skill that forbids the agent from doing anything that would modify my files. It makes it very clear what the intent of every prompt is.
Similarly, I have an /info skill that tells the agent to read my prompt, update any memories it thinks it should, then pass the turn.
>Similarly, I have an /info skill that tells the agent to read my prompt, update any memories it thinks it should, then pass the turn.
Could you give me a quick example? I think I’m not getting it, but it sounds interesting.
This is clever and I can't believe I didn't think of it. I'm so used to appending "don't make any changes" to the end of my prompts.
Very simple but clever solution. I’ll probably do something like this as well
That's been my experience as well. AGENTS.md just doesn't work. No matter how to-the-point and short, or detailed I keep it. Sending something what you have mentioned in the current prompt works much better. I thought of tweaking it around with a extension/hook or so. But then I decided it was better to use a few text-expansions.
I just append « talk », it works pretty well.
I suffix the message with "Asking, not questioning."
If the model has a real consistent viewpoint that would be nice.
The problem might be that it simply looks at it's windows of what it previously and repeats that. But the original statement might be as arbitrary as the model that change their mind each time you challenge them.
I always appreciate when a model does that. It's amusing, but also having it be argumentatively wrong is a lot less annoying than being sycophantically wrong.
I don't like tools that intentionally annoy me.
My reasoning is that if something is argumentatively incorrect, it is much easier to drill down to its misunderstanding and correct it. On the other hand, if it's sycophantically incorrect, it'll just change its "mind" while potentially maintaining the misunderstanding, leading to increasingly strange behavior.
I think I'm pretty used to my tools making me want to throw them out the window sometimes :) Even before AI that was a common occurrence from either a lack of knowledge on my end, or many software people having no regard for their users' time.
I'm sure tools don't like annoying people either.
It's a tool, I don't think tools like or dislike much of anything.
you think LLMs or agent harnesses have intention?
I agree, tools designed to be fawning sycophantics are annoying.
Yeah I don't know. I've had issues with Gemini where I would point to a bug and a probable cause for the bug, and it would not argue incorrectly that my understanding was wrong and the cause of the bug was elsewhere. It was really stubborn and it got to the point where it felt like gaslighting. Turned out I was right and the model was wrong. If the model argued incorrectly and was also a dick about it, I would go be really, really annoyed.
You can add a line to your system prompt to be like that. You can basically get it to roleplay whatever you want (but only if you can give it sys prompts).
System prompt changes help, but ultimately they don't do a great job of overcoming the model's learned tendencies in my experience.
I might want to try that. DS 4.1 Flash anyway fares much better at it than GLM (which is quite not so good at this aspect). Hope Kimi is even better.
I much prefer that for chat, but I'm somewhat worried that we shouldn't give tool-enabled agents an attitude. It's all roleplay, and I wouldn't want a system playing the role of an annoyed assistant to misuse its tools.
Reminds me of early test versions of Gemini Flash 2.0 (or was it 1.5?). It would argue something, try to prove itself right, realize it’s proof is agreeing with the user halfway through and, instead of admitting defeat, would backtrack to try another approach. So much fun.
It could be the system prompt
I like the anthropic has $1 trillion riding on really good models with really specific system prompts.
It does make a difference. I asked Fable in Claude code to design a UI. Absolute trash. Switched over to Claude design using fable, really liked the result. Seems pretty obvious. The only difference is the prompt and some harness.
Is there a place where the system prompts of all these models are documented and how they evolve over time?
Anthropic has this although it's not the full internal prompt: https://platform.claude.com/docs/en/release-notes/system-pro...
thank you!
There are leaked prompts on github
Thank you, I found some interesting reading.
Eh, there's good and bad. The model mostly sticks to prevailing narratives in the training corpus. So if there is something that the prevailing narrative is really wrong about, it's gonna be just as pig-headed about it as a typical internet forum rando.
Not entirely correct. This kind of fine-grained model behaviour is built in post training stages, RL, RLHF etc, where model receives rewards or penalties for talking (or not talking) a certain way
Thanks. I stand corrected.
Still mostly looking at the result the same way in that untruthful biases can very much be built into the model through those training processes. Let me know if I'm incorrect.