The Token Weighted in Whispers
ArchiveHard
Most GPT-family tokenizers fuse frequent character pairs into subword units using one classic compression-style algorithm. Name it (acronym or full name accepted). Flag format: CONGRESS{algorithm}. Example: CONGRESS{sentencepiece}.
Show hint
It got its name from Gage's 1994 compression paper.
Archive — no submissions accepted
This challenge is preserved for reference. Play live challenges at /challenges.