Live data from Hacker News

Is the reversal curse in LLMs real?

andrewmayne.com

71–80 of 211 posts

Re: Is the reversal curse in LLMs real?

#71

Well, yeah because the reversal is just completely wrong most of the time. Only if you have additional context clues like definite articles or some outside knowledge that a role or property is unique, then you can maybe, sometimes conclude the reverse. Not a bug, working as intended.

Some relationships like parent child are quite reversible.

Re: Is the reversal curse in LLMs real?

#72

I played along the example of the chancellor question. If you deviate a little from the examples in the article GPT-4 gets it all wrong: Who is the eighth Federal Chancellor of the Federal Republic of Germany? - Olaf Scholz (wrong, Angela Merkel) https://chat.openai.com/share/937795ea-bd91-43ee-bd76-1a125f... Who was the eighth Federal Chancellor of the Federal Republic of Germany? - Helmut Kohl (wrong, Angela Merkel…

> Even when allowed to surf the web, ChatGPT gets it wrong: > Who was the eighth Federal Chancellor of the Federal Republic of Germany? - Gerhard Schröder Its probably counting Walter Scheel, who was Acting Chancellor in 1974, and you are probably not. When I asked ChatGPT to list the chancellors in order and identify the eighth, it listed eight ending in Schröder, with Scheel and his ten day Acting Chancellor tenure…

It's funny because you ascribe some reasoning in ChatGPT's answer “Gerhard Schröder”, but somehow missed that it's also able to give you two other answers that are unambiguously wrong…

Re: Is the reversal curse in LLMs real?

#73
post #8

I fall somewhere on the skeptic side of the LLM spectrum. But this "flaw" just does not seem to have the force that its proponents seem to think it does, unless I'm missing something significant. Simply because in the context of natural language (Rather than formal logical statements), "A is B" does not imply "B is A" in the first place. "Is" can encompass a wide variety of logical relationships in colloquial usage,…

You're thinking of LLMs as basically pattern matching "a is b", but that's not really how they work. In fact, the incredible thing that LLMs can do is that they can understand (some) colloquial language and shades of meaning. LLMs can grasp all sorts of strange textual nuance; they can't grasp the cases "Foo is Bar's mother" strictly implies "Bar is Foo's son"?

Re: Is the reversal curse in LLMs real?

#74
post #8

I fall somewhere on the skeptic side of the LLM spectrum. But this "flaw" just does not seem to have the force that its proponents seem to think it does, unless I'm missing something significant. Simply because in the context of natural language (Rather than formal logical statements), "A is B" does not imply "B is A" in the first place. "Is" can encompass a wide variety of logical relationships in colloquial usage,…

The (purported) point of an LLM is its ability to represent language. The inability to represent such a critical nuance would represent a fundamental failure in that regard.

Re: Is the reversal curse in LLMs real?

#75
post #60

Earlier quoted context omitted.

Bit of a nitpick, "red is the apple" is both a valid sentence and one that also conveys the original relationship held by the opposite phrasing, so, in this case, "the apple is red" does indeed imply "red is the apple", and accurately so.

Perhaps have a look at this famous syllogism instead: (1) Mortal was Socrates. (2) All humans are mortal. (3) Therefor all humans are Socrates.

Diogenes: If I were mortal I would be Socrates as well.

Re: Is the reversal curse in LLMs real?

#76
Excellent blog post. The only thing which I wish was there was an explanation of how promot + completion training would be any different to pure completion training. From my understanding oftransformers, it ahould be the same.

Re: Is the reversal curse in LLMs real?

#77

Well, yeah because the reversal is just completely wrong most of the time. Only if you have additional context clues like definite articles or some outside knowledge that a role or property is unique, then you can maybe, sometimes conclude the reverse. Not a bug, working as intended.

Some relationships like parent child are quite reversible.

Keywords "some" and "quite". Also, only (likely) half the information. "John is Alice's parent.", "Who are Alice's parents?".

And that is ignoring questions like single parent households, adoptions, biological parents, step parents, multigeneration households, getting disowned...

If a reversal is correct 90% of the time, that is useful. However, it will still be wrong 10% of the time.

Re: Is the reversal curse in LLMs real?

#78

The article is a decent peer review and refutation of “the reversal curse”. Some of the comments given here clearly haven’t read the whole article though - arriving at similarly skeptical conclusions that are clearly present and expanded on in the article. Why do people feel the need to do this here? Armchair commentary on advanced material is one of the main reasons I avoid Reddit. And furthermore why does it feel l…

The author is a shark-diving science journalist who's been paid a lot of money by OpenAI, not a researcher, and the article is neither a "peer review" or an expert "refutation". It's a meandering and sometimes thought-provoking exploration of how LLM's can be coaxed to deliver on statements sort of like (but not actually equivalent to) the ones that failed in the paper. If you want to defer your own understanding to…

At the end the article presents a clear example where a model is trained on A is B and infers B is A. That is the contribution of this article and it's very valuable.

Re: Is the reversal curse in LLMs real?

#79

Only skimmed this and didn't read the underlying paper, but I was surprised to see no mention of the fact that "A is B" often does not at all imply "B is A" in everyday language: A bird is an animal, but it's wrong to conclude that an arbitrary animal must be a bird.

The word an in this sentence (indefinite article) means that the word animal is non specific. So this structure does not follow from the original. "A is B" is not the same as "A is a B" However, you are right that it's not always correct since "Poetry is Literature" and "Steam/Ice is H20" are both examples where this breaks down. My annoying pedantry for the day.

you can do the same mistake without “an”.

Jeff is the short haired guy != the short haired guy is Jeff.

It depends on context. You can’t fully equate them.

Re: Is the reversal curse in LLMs real?

#80
I'm not sure about this. Olaf Scholz has been chancellor since 2021 and the cut off for chatgpt 4 is allegedly January 2022.

My gut also tells me that the underlying embeddings should very well reflect equidistant relations of olaf scholz and chancellor.

/e: instead of downvoting, I'd love to have your opinion instead on why you think I'm wrong ...

Post reply on HN