MDM-Prime-v2: Binary Encoding and Index Shuffling Enable Scaling of Diffusion Language Models
Open paper
Benchmark
Model
Rank
Results
social-commonsense-reasoning-on-siqa
MDM-Prime-v2 (1.1B)
#12
Accuracy: 42.02
common-sense-reasoning-on-arc-easy
MDM-Prime-v2 (1.1B)
#82
Accuracy: 47.81
question-answering-on-boolq
MDM-Prime-v2 (1.1B)
#88
Accuracy: 62.05
Rank counts only results with a code link.