List your company

Language Modeling on Paloma-C4 (val)

58.48PerplexityPartially Reverse BDLM Mamba-H
57.489264.177170.86577.5529Jul 2, 2026

Evaluation results

MethodLinks
Partially Reverse BDLM Mamba-HLR=8×10−32026.0758.481.28
BDLM attentionLR=4×10−32026.0770.811.34
Full-sequence DiffuMamba-HLR=4×10−32026.0780.421.38
Full-sequence attentionLR=8×10−32026.0783.251.39