One particular thing, unrelated to the linguistic argument itself, stood out to me. In the PCA visualisation, we can see that some sequences of layers have particularly tight and stationary clusters. Incidentally, those are also exactly the layers that the previous RYS post identified as most useful to repeat to improve perfomance on the probes.
I wonder, if that correlation could be used to identify good candidates for repeating layers.
I don't care too much about the article being written with LLM support. There is actual work being done that is being showcased here. I'd rather read an LLM version of it, rather that not learning about those things at all.
One thing to benchmark is if LLMs are better at solving complex problems if they're described in one language vs others.
There's SWE-bench Multilingual for example, but translating a problem into multiple natural languages before passing it to the LLM has not been benchmarked afaik.
If there's some residual of the natural language left when the middle layers execute, that would in part validate Sapir-Whorf.
The linguistic argument is fascinating.
One particular thing, unrelated to the linguistic argument itself, stood out to me. In the PCA visualisation, we can see that some sequences of layers have particularly tight and stationary clusters. Incidentally, those are also exactly the layers that the previous RYS post identified as most useful to repeat to improve perfomance on the probes.
I wonder, if that correlation could be used to identify good candidates for repeating layers.
"This isn’t just correlation. It’s a complete structural reorganisation of the representation space."
I don't care too much about the article being written with LLM support. There is actual work being done that is being showcased here. I'd rather read an LLM version of it, rather that not learning about those things at all.
One thing to benchmark is if LLMs are better at solving complex problems if they're described in one language vs others.
There's SWE-bench Multilingual for example, but translating a problem into multiple natural languages before passing it to the LLM has not been benchmarked afaik.
If there's some residual of the natural language left when the middle layers execute, that would in part validate Sapir-Whorf.