The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" (2023)
1–10 of 54 posts
Re: The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" (2023)
#2[dead]
Re: The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" (2023)
#3I remember seeing this popping up in discussions the first time, but never noticed any resolution (other than to train both sides). Has SOTA advanced?
Re: The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" (2023)
#4(2023)
Re: The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" (2023)
#5As I and several other people pointed out last time this was posted, "A is B," in natural language, does not imply "B is A." "Is" can denote any of many different shades of relationship weaker than logical identity.
Re: The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" (2023)
#6“A is B” doesn’t generally entail “B is A”.
“A square is a rectangle” does not entail “a rectangle is a square”.
Similarly, “Socrates is alive” doesn’t entail “alive is Socrates”.
Notably, they mention when context is included, LLM performance rises — ie, exactly when we include extra information that allows it to recognize what kind of information is being conveyed.
But the LLM is correct not to generalize that pattern when it doesn’t generalize — even if researchers have salient example, but ignore contrary ones (eg, square-rectangle or Socrates-alive).
Re: The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" (2023)
#7As I and several other people pointed out last time this was posted, "A is B," in natural language, does not imply "B is A." "Is" can denote any of many different shades of relationship weaker than logical identity.
Even in strict logic, “is” can denote membership, as in, “squares are rectangles” does not entail “rectangles are squares”.
Re: The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" (2023)
#8As I and several other people pointed out last time this was posted, "A is B," in natural language, does not imply "B is A." "Is" can denote any of many different shades of relationship weaker than logical identity.
Yes, "Who is Mary Lee Pfeiffer's son?" happens to have just one answer, whereas "Who is Mary Lee Pfeiffer's child?" would have several.
Re: The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" (2023)
#9Not only is the inverse not generally true (as others have pointed out), their examples requires several mental leaps.
"Who is Tom Cruise's mother? [A: Mary Lee Pfeiffer]" and the reverse "Who is Mary Lee Pfeiffer's son?"
The word "mother" has no relationship to "son" in terms of the model, and so while the model might be able to infer a proximity relationship between "Tom Cruise" and "Mary Lee Pfeiffer" just because they appear in the same sentence, expecting the AI to guess that the inverse of mother is son is a bit of a stretch, especially when they're both lossy mappings, because the relationship is {mother,father} {son,daughter}. If we're going to train models to make that mental leap, we'd have to put up with false results like "Tom Cruise is the daughter of Mary Lee Pfeiffer" unless the model is also supposed to infer that Tom means he can only be a son.
Re: The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A" (2023)
#10Here's a short example from A. Karpathy in a 2024 video: https://www.youtube.com/watch?v=zjkBMFhNj_g&t=750s