3 ms·
> If you rarely got to see letters and just saw fragments of words as something like Chinese characters (tokens), could you count the R's in arbitrary words wel
by ben_w 1y ago
> If you rarely got to see letters and just saw fragments of words as something like Chinese characters (tokens), could you count the R's in arbitrary words well?
While this seems correct, I'm sure I tried this when it was novel and observed that it could split the word into separate letters and then still count them wrong, which suggested something weird is happening internally.
I just now tried to repeat this, and it now counts the "r"'s in "strawberry" correctly (presumably enough examples of this specifically on the internet now?), but I did find it making the equivalent mistake with a German word (https://chatgpt.com/share/6859289d-f56c-8011-b253-eccd3ceceea7 https://chatgpt.com/share/6859289d-f56c-8011-b253-eccd3cecee...):
How many "n"'s are in "Brennnessel"?
But even then, having it spell the word out first, fixed it: https://chatgpt.com/share/685928bc-be58-8011-9a15-44886bb5225a https://chatgpt.com/share/685928bc-be58-8011-9a15-44886bb522...
- kbelder 1y agoCounting letters is such a dull test. LLMs generally have a hard time with this question because letters are tokenized before they receive them, and they have to go through an involved reasoning process to figure it out. It's like asking a color blind person what color the street light is, and declaring him unintelligent because he sometimes gets the answer wrong.
- moomin 1y agoI mean, if you don’t want to include tests that LLMs are, by definition, bad at, why don’t we do the same thing for humans?
- cma 1y agoBecause this is a gotcha test and they are trained on a different tokenized transform of the data that makes it hard. As a system they can output it as a string and have an interpreter, as part of the system, count the letters. So no real meaningful deficiency except they maybe don't waste system prompt space to tell it to do that, as a choice.