The Four-Bit Adapter
ArchiveVery Easy
Dettmers et al. (2023) combined low-rank adaptation with 4-bit NF4 quantization of the frozen base to let 65B models fine-tune on a single 48GB GPU. Which name?
Show hint
The previous method + a letter signifying precision.
Archive — no submissions accepted
This challenge is preserved for reference. Play live challenges at /challenges.