How outdated information hides in LLM token generation probabilities
1–10 of 55 posts
Re: How outdated information hides in LLM token generation probabilities
#2> The scenario that I’m worried about, and that is playing out right now, is that they get good enough that we (or our leaders) become overconfident in their abilities and start integrating them into applications that they just aren’t ready for without a proper understanding of their limitations.
Very true.
Re: How outdated information hides in LLM token generation probabilities
#3The o1 example is interesting. In the CoT summary it acknowledges that the most recent official information is 1611m, but it then chooses to say 1622 because it's more commonly cited. It's like it over-thinks itself into the wrong answer.
Re: How outdated information hides in LLM token generation probabilities
#4The o1 example is interesting. In the CoT summary it acknowledges that the most recent official information is 1611m, but it then chooses to say 1622 because it's more commonly cited. It's like it over-thinks itself into the wrong answer.
Does it search the internet for that? I assume so because else claiming how often something is cited does not make sense, but would be interesting to know surely. Even gpt4o mini with kagi gets it right with search enabled (and wrong without search enabled - tried over a few times to make sure).
Re: How outdated information hides in LLM token generation probabilities
#5The o1 example is interesting. In the CoT summary it acknowledges that the most recent official information is 1611m, but it then chooses to say 1622 because it's more commonly cited. It's like it over-thinks itself into the wrong answer.
Does it search the internet for that? I assume so because else claiming how often something is cited does not make sense, but would be interesting to know surely. Even gpt4o mini with kagi gets it right with search enabled (and wrong without search enabled - tried over a few times to make sure).
Can the claim about citation frequency be just an answer pattern and not model's exact reasoning?
Re: How outdated information hides in LLM token generation probabilities
#6Earlier quoted context omitted.
Does it search the internet for that? I assume so because else claiming how often something is cited does not make sense, but would be interesting to know surely. Even gpt4o mini with kagi gets it right with search enabled (and wrong without search enabled - tried over a few times to make sure).
Can the claim about citation frequency be just an answer pattern and not model's exact reasoning?
Yeah there could be parts of the training set with 1611 being explicitly called the official and 1622 being explicitly called the most common answer. But it can also have access to search results directly I think. Is there a way to know if it does or not?
Re: How outdated information hides in LLM token generation probabilities
#7> Welcome to the era of generative AI, where a mountain can have multiple heights, but also only one height, and the balance of my bank account gets to determine which one that is. All invisible to the end user and then rationalised away as a coincidence.
I've always found the idea of untraceable, unfixable, unpredictable bugs in software... Offensive. Dirty. Unprofessional.
So the last couple years have been been disconcerting, as a non-trivial portion of people who I thought felt similarly started to overlook it in LLMs, while also integrating those LLMs into flows where the bad-output can't even be detected.
Re: How outdated information hides in LLM token generation probabilities
#8The o1 example is interesting. In the CoT summary it acknowledges that the most recent official information is 1611m, but it then chooses to say 1622 because it's more commonly cited. It's like it over-thinks itself into the wrong answer.
Does it search the internet for that? I assume so because else claiming how often something is cited does not make sense, but would be interesting to know surely. Even gpt4o mini with kagi gets it right with search enabled (and wrong without search enabled - tried over a few times to make sure).
I don’t think the public o1 can search the internet yet, unlike 4o. In principle it could know that something is more commonly cited based on its training data. But it could also just be hallucinating.
Re: How outdated information hides in LLM token generation probabilities
#9The o1 example is interesting. In the CoT summary it acknowledges that the most recent official information is 1611m, but it then chooses to say 1622 because it's more commonly cited. It's like it over-thinks itself into the wrong answer.
How could a language model infer that the official information overrules anything else?
Re: How outdated information hides in LLM token generation probabilities
#10The o1 example is interesting. In the CoT summary it acknowledges that the most recent official information is 1611m, but it then chooses to say 1622 because it's more commonly cited. It's like it over-thinks itself into the wrong answer.
How could a language model infer that the official information overrules anything else?
I’m not sure what kind of response you’re looking for, or if this is a rhetorical question or not. But “how could a language model infer…?” can be asked about a whole lot of things that language models have no problem reliably inferring.