this post was submitted on 03 Mar 2026
62 points (97.0% liked)
Health - Resources and discussion for everything health-related
4261 readers
248 users here now
Health: physical and mental, individual and public.
Discussions, issues, resources, news, everything.
See the pinned post for a long list of other communities dedicated to health or specific diagnoses. The list is continuously updated.
Nothing here shall be taken as medical or any other kind of professional advice.
Commercial advertising is considered spam and not allowed. If you're not sure, contact mods to ask beforehand.
Linked videos without original description context by OP to initiate healthy, constructive discussions will be removed.
Regular rules of lemmy.world apply. Be civil.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I don’t think you do either, or at least you don’t understand LLMs.
LLMs are absolutely deterministic but to make them useful we use sampling to focus or diversify results, and adjust the probability distribution etc to control variability/randomness.
Without those variables (top-p and temperature, etc), transformers are indeed deterministic.
There’s other aspects that influence such as floating point optimisations, hardware latency etc. But the main reason you don’t encounter deterministic LLMs day to day is because they’re built not to be. But the core of what makes an LLM an LLM is technically deterministic.
I mean, that's kinda like saying a random number generator can be deterministic. It can be, but that's not how it's used.
Sure LLMs can be deterministic, but they aren't in practice cause it makes the results worse. If you prompt any production LLM with the same inputs, you aren't guaranteed the same outputs.