3 ms·
The chat context isn't as big as is available in the API - was the author name in its context window? Papers can be fairly long
by IanCal 4y ago
The chat context isn't as big as is available in the API - was the author name in its context window? Papers can be fairly long
- dfox 4y agoApparently the model is intentionally trained such that it does not “know” any kind of metadata about its training inputs. So it can deduce the name of the author if it was referenced elsewhere, but not from the paper itself.
- toss1 4y agoThat's interesting, but why would the data we provide for context or explicitly for summarization be treated like training data? And if it is excluded, why wouldn't it just default to a response resembling "I am not setup to identify and extract that data."?
- toss1 4y agoGood thought, but the paper was a couple of pages long, max. Less than 2500 words, less than 15K characters (sorry, I should have mentioned that). Also, the fact that it could pick up the provided FName and still hallucinate MidName and Surname and Employer kind of gives away that it was in context. It is really funny (as long as it isn't a serious situation) how it cannot identify anything, but the context will give it cues to hallucinate