Other recent work focuses on using the “modeling” component from well-known compressors to do
other tasks. [34] uses the model from GZip to perform text classification. [68 ] uses the Arithmetic
Decoding algorithm with an LLM as the model to do diverse parallel sampling from that LLM. One
could imagine that the “model” of our compressors (M1) is a teacher for M2, but unlike these...
(Optional:) Please add any files, screenshots, or other information here.
No response
(Required) What is this issue most closely related to? Select one.
Description
ends at https://arxiv.org/pdf/2404.03626.pdf#page=6
(Optional:) Please add any files, screenshots, or other information here.
No response
(Required) What is this issue most closely related to? Select one.
Other
Internal issue ID
NA
Paper URL
https://arxiv.org/html/2404.03626v1
Browser
Firefox
Device Type
desktop