elcomet

That's not my experience, the model is adding tons of unnecessary details, that make it much harder to find the important information for an expert that already knows all those details.

I don't want ELI5, I just want the model to go straight to the point, like a colleague human expert would do.

Why do you assume this order? I would assume the AI starts with lean (that's what it was RL'd on at least) then tried to translate the proof for us humans, which is quite hard.

I'm not sure about it but it seems plausible given the potential error.

Edit: they explicite say that your order is correct in the post

Why not using a cheap LLM with thinking completely disabled ? I don't think it will be much more expensive than jev.

Strange they don't mention the Pink Floyd album in the article!

Hallucinations were defined in the context of text generation models so your question does not really make sense.

IMO your system can make mistakes that are similar in spirit to hallucination (i.e. answering with a false answer instead of abstaining to answer).

The technology and the results are very handwavy. What is RLCD exactly ? What are scores on benchmarks compared to LLMs ?

This website does not inspire confidence at all, it all sounds like a marketing piece. I wish it was true, some kind of text-prompted classifier with LLM performance would be cool, but I can't trust it with what we are given.

We didn't have the compute required (GPUs powerful enough to parallelize forward and backward pass). This compute is what allows us to train from human knowledge or distillation.

UUID as key is useful when you have a distributed system where multiple workers create items independently

Could you share your config?

The heat pump will pull heat from inside the house? This sounds terrible for efficiency in winter, as you will need to reheat the room

Yeah this sounds disingenuous. I can also see my heater to 70C if I want, that does not increase its size...

I don't understand all the hype for generating SVG with LLM. The task is not really useful, doesn't seem that interesting in single shot as it's really hard, and no human could do it (it would be more useful if the model has visual feedback and could correct the result).

And also, since it becomes a popular task, companies will add the examples in their training set, so you're just benchmarking who has the better text to SVG training set, not the overall quality of the model.

I'd say that thoughts and reasoning are two different things, you're moving the goalpost.

But what makes the computer hardware fundamentally incompatible with thinking? Compared to a brain

How can you know?

Can you explain more? Which things are impossible in blender

Too many acronyms, what's FE, BFF?

Arc browser unifies the tabs and bookmarks in a very clever way.

I'm wondering if you can prompt it to work like this - make minimal changes, and run the tests at each step to make sure the code is still working

If it's chromium based, they will need to remove manifest v2 at some point to stay close to the upstream version.

You can offload tensors to the cpu memory. It will make your model run much slower but it will work

I can't tell if this is a joke if if you're serious

Book clubs, art clubs, movie club... Lot of options

What do you mean? Async/await uses threads

Multi-step reasoning means that the LLM is giving a question (maths here), and generating an answer that consists of many intermediate words, before returning the solution. Here, we don't want to tell the LLM how to solve the problem word-by-word. We want to tell it at the end, "correct" or "incorrect", and have the model learn on its own to generate intermediate steps, to reach the solution.

That's typically a setup where RL is desirable (even necessary): we have sparse rewards (only at the end) and give no details to the model on how to reach the solution. It's similar to training models to play chess against a specific opponent.

He means 10 humans voting for the answer

Those are tree search techniques, they are not metrics to assess the "human" complexity of a line. They could be used for this purpose but out of the box they just give you winning probability

I'm wondering why there isn't a cheapest alternative to the pi without those features that most find useless. Or a more powerful for the same price

It's possible, the question is how to choose which submodel will be used for a given query.

You can use a specific LLM, or a general larger LLM to do this routing.

Also, some work suggest using smaller llms to generate multiple responses and use a stronger and larger model to rank the responses (which is much more efficient than generating them)

Aren't you afraid of being now stuck with GCP?