List your company
ResearchFormerly WizWand

Text Compression on Wiki-40B

214Compressed Size (MB)R-L1
166.16489.088121,134.92May 23, 2025
Updated 5mo ago

Evaluation results

MethodLinks
R-L1Number of parameters=308M, Regularization type=R-L1, Coding scheme=Arithmetic coding, Chunk size=512 bytes2025.05214
DRRNumber of parameters=308M, Regularization type=DRR, Coding scheme=Arithmetic coding, Chunk size=512 bytes2025.05239
UT4Number of parameters=4M, Regularization type=unregularized, Coding scheme=Arithmetic coding, Chunk size=512 bytes2025.05254
PMMPNumber of parameters=308M, Regularization type=PMMP, Coding scheme=Arithmetic coding, Chunk size=512 bytes2025.05299
UT25Number of parameters=25M, Regularization type=unregularized, Coding scheme=Arithmetic coding, Chunk size=512 bytes2025.05306
ZSTDtype=conventional compressor2025.05802
LZMA2type=conventional compressor2025.05992
UT308Number of parameters=308M, Regularization type=unregularized, Coding scheme=Arithmetic coding, Chunk size=512 bytes2025.051,410