List your company

Multiple-choice Question Answering on Winogrande (test)

63.8AccuracyPPT (Distribution)
46.53651.01855.559.982Feb 24, 2026Mar 11, 2026Mar 26, 2026Apr 10, 2026Apr 25, 2026May 10, 2026May 26, 2026

Evaluation results

MethodLinks
PPT (Distribution)Backbone=QWEN2-7B2026.0563.8
BaseBackbone=QWEN2-7B2026.0561.3
PPT (Mean)Backbone=QWEN2-7B2026.0561.1
PPT (Distribution)Backbone=LLAMA3-8B2026.0558.1
BaseBackbone=LLAMA3-8B2026.0554.5
PPT (Mean)Backbone=LLAMA3-8B2026.0554.2
AR Transformer2026.0252.8
MDLM2026.0251.93
VerbalizedBackbone=LLAMA3-8B2026.0550.7
Duo++k=52026.0250.12
Duo++k=22026.0249.64
Duo++k=32026.0247.36
Duo2026.0247.2