3 ms·
You can easily demonstrate that an LLM does know certain fact X AND demonstrate that the LLM will deny that they know fact X (or be flaky about it, randomly de
by markwkw 2y ago
You can easily demonstrate that an LLM does know certain fact X
AND demonstrate that the LLM will deny that they know fact X (or be flaky about it, randomly denying and divulging the fact)
There are two explanations:
A. They lack self-reflection
B. They know they know fact X, but avoid acknowledging for ... reasons?
I find the argument for A quite compelling
- throwaway290 2y agoThey will say "it's just a bad LLM", don't bother
- astrange 2y ago> demonstrate that the LLM will deny that they know fact X (or be flaky about it, randomly denying and divulging the fact) No, the sampling algorithm you used to query the LLM does that. Not the model itself. e.g. https://arxiv.org/pdf/2306.03341.pdf https://arxiv.org/pdf/2306.03341.pdf > B. They know they know fact X, but avoid acknowledging for ... reasons? That reason being that the sampling algorithm didn't successfully sample the answer.