The Ill-Posed Curse Of Dimensionality
ArchiveHard
A 2023 embedding technique trains a single model such that any prefix (first 256, 512, 1024, etc. dimensions) is itself a valid, usable embedding — allowing variable-precision retrieval. Name the technique (one word, after a famous doll). Flag format: CONGRESS{name}. Example: CONGRESS{onionnet}.
Show hint
A Russian toy; layers within layers.
Archive — no submissions accepted
This challenge is preserved for reference. Play live challenges at /challenges.