Variable-Width Transformers
Open paper
Benchmark
Model
Rank
Results
reading-comprehension-on-race
3B/1B MoE > <former
#31
Accuracy: 34.3
question-answering-on-boolq
2B > <former
#96
Accuracy: 60.9
common-sense-reasoning-on-winogrande
2B > <former
#116
Accuracy: 60.2
Rank counts only results with a code link.