Language Modeling on WikiText (val)

12.51Perplexity

MDM-Prime-v2

Updated 25d ago

Evaluation Results

Method	Links
MDM-Prime-v2 2026.03		12.51
Block Diffusion + Gumbel Distillation 2026.03		13.86
MDLM + Gumbel Distillation 2026.03		15.57
MDM-Prime 2026.03		19.46
AsyncMesh 2026.01		21.14
FullSync 2026.01		21.23
DPAvg 2026.01		21.26
SPARTA 2026.01		21.3
ARM 2026.03		22.28
MDM-Prime 2026.03		23.86
AsyncSPARTA 2026.01		24.8
ARM 2026.03		25.75
MDM-Prime 2026.03		25.77
MDM-Prime-v2 2026.03		26.05
AR Transformer 2025.06		26.54
MDM 2026.03		26.65
MDM-Prime 2026.03		26.87
MDM-Prime 2026.03		27.93
EDLM-coAR 2026.03		28.31
EDLM-NCE 2026.03		30.77
Block Diffusion 2026.03		30.82
MASA-QKV 2025.08		30.83
Transformer-L 2025.08		30.88
BD3-LM 2026.03		31.31
MASA-QKVO 2025.08		31.34
GQA 2025.08		31.74
Repeat-all-over 2025.08		32.27
Seq-Sharing 2025.08		32.43
MDM 2026.03		32.83
BD3-LM 2025.06		32.88
Low-Rank 2025.08		33.28
Duo 2026.03		33.57
SEDD 2026.03		34.28
MDLM 2026.03		34.52
Eso-LMs 2025.06		35.65
MDLM 2025.06		37.08
Eso-LMs 2025.06		37.32
SEDD Absorb 2025.06		38.55
Eso-LMs 2025.06		39.57
MASA-QKV 2025.08		42.31
Transformer-M 2025.08		44.49
MASA-QKVO 2025.08		45
Eso-LMs 2025.06		45.08
GQA 2025.08		46.21
Seq-Sharing 2025.08		47.36
Low-Rank 2025.08		47.48
Repeat-all-over 2025.08		47.63
MASA-QKV 2025.08		72.08
MASA-QKVO 2025.08		72.82
Transformer-S 2025.08		76.11
GQA 2025.08		78.41
Repeat-all-over 2025.08		78.97
Seq-Sharing 2025.08		80.35
Low-Rank 2025.08		83.25