zlacker

>>zdw+(OP)
The problem with viewing LLMs as just sequence generators, and malbehaviour as bad sequences, is that it simplifies too much. LLMs have hidden state not necessarily directly reflected in the tokens being produced and it is possible for LLMs to output tokens in opposition to this hidden state to achieve longer term outcomes (or predictions, if you prefer).

Is it too anthropomorphic to say that this is a lie? To say that the hidden state and its long term predictions amount to a kind of goal? Maybe it is. But we then need a bunch of new words which have almost 1:1 correspondence to concepts from human agency and behavior to describe the processes that LLMs simulate to minimize prediction loss.

Reasoning by analogy is always shaky. It probably wouldn't be so bad to do so. But it would also amount to impenetrable jargon. It would be an uphill struggle to promulgate.

Instead, we use the anthropomorphic terminology, and then find ways to classify LLM behavior in human concept space. They are very defective humans, so it's still a bit misleading, but at least jargon is reduced.

>>barrke+k5
Do they ? LLM embedd the token sequence N^{L} to R^{LxD}, we have some attention and the output is also R^{LxD}, then we apply a projection to the vocabulary and we get R^{LxV} we get therefore for each token a likelihood over the voc. In the attention, you can have Multi Head attention (or whatever version is fancy: GQA,MLA) and therefore multiple representation, but it is always tied to a token. I would argue that there is no hidden state independant of a token.

Whereas LSTM, or structured state space for example have a state that is updated and not tied to a specific item in the sequence.

I would argue that his text is easily understandable except for the notation of the function, explaining that you can compute a probability based on previous words is understandable by everyone without having to resort to anthropomorphic terminology

>>d3m0t3+J8
There is hidden state as plain as day merely in the fact that logits for token prediction exist. The selected token doesn't give you information about how probable other tokens were. That information, that state which is recalculated in autoregression, is hidden. It's not exposed. You can't see it in the text produced by the model.

There is plenty of state not visible when an LLM starts a sentence that only becomes somewhat visible when it completes the sentence. The LLM has a plan, if you will, for how the sentence might end, and you don't get to see an instance of that plan unless you run autoregression far enough to get those tokens.

Similarly, it has a plan for paragraphs, for whole responses, for interactive dialogues, plans that include likely responses by the user.

>>barrke+S9
The LLM does not "have" a plan.

Arguably there's reason to believe it comes up with a plan when it is computing token propabilities, but it does not store it between tokens. I.e. it doesn't possess or "have" it. It simply comes up with a plan, emits a token, and entirely throws all its intermediate thoughts (including any plan) to start again from scratch on the next token.

>>gpm+Xj
This is wrong, intermediate activations are preserved when going forward.

>>lostms+nq
Within a single forward pass, but not from one emitted token to another.

>>ACCoun+GT
What? No. The intermediate hidden states are preserved from one token to another. A token that is 100k tokens into the future will be able to look into the information of the present token's hidden state through the attention mechanism. This is why the KV cache is so big.

>>andy12+wB1
KV cache is just that: a cache.

The inference logic of an LLM remains the same. There is no difference in outcomes between recalculating everything and caching. The only difference is in the amount of memory and computation required to do it.

>>ACCoun+XB3
The same can be said about any recurrent network. To predict the token n+1 you could recalculate the hidden state up to token n, or reuse the hidden state of token n from the previous forward pass. The only difference is the amount of memory and computation.

The thing is that, fundamentally, an auto-regressive transformer is a model whose state grows linearly with each token without compression, which is what bestows them with (theoretical) perfect recall.