Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Commonsense Reasoning on WinoG (test val)
Loading...
73.88
Accuracy
LLaDA (Base)
54.9936
59.8968
64.8
69.7032
Feb 19, 2026
Accuracy
Updated 4d ago
Evaluation Results
Method
Method
Links
Accuracy
LLaDA (Base)
Pruning Ratio=0.0
2026.02
73.88
Sink-Aware (Ours)
Pruning Ratio=0.3
2026.02
67.4
LLaDA-structure
Pruning Ratio=0.3
2026.02
66.3
Sink-Aware (Ours)
Pruning Ratio=0.5
2026.02
57.24
LLaDA-structure
Pruning Ratio=0.5
2026.02
55.72
Feedback
Search any
task
Search any
task