SuperBPE: Space Travel for Language Models
Paper
•
2503.13423
•
Published
•
13
SuperBPE
SuperBPE tokenizers and models trained with them
-
-
UW/OLMo2-8B-SuperBPE-t180k
Text Generation • 8B • Updated • 45 • 8 -
UW/OLMo2-11B-SuperBPE-t180k
Text Generation • 11B • Updated • 5 • 2 -
UW/OLMo2-8B-SuperBPE-t160k
Text Generation • 8B • Updated • 6 • 3 -
UW/OLMo2-8B-SuperBPE-t80k
Text Generation • 8B • Updated • 1 -
UW/OLMo2-8B-BPE
Text Generation • 8B • Updated • 2 -
6
SuperBPE Playground
📝Experiment with and compare different tokenizers
-
UW/olmo-mix-1124-subset-p99
Updated • 245 • 1