List your company

Visual Reasoning on NLVR2 v1 (test/val)

83.8AccuracyCVP
82.86483.10783.3583.593Jun 15, 2026

Evaluation results

MethodLinks
CVPBackbone=BEiT-3-base, Rel. Time=2.5×2026.0683.85.2414.623.50.36—1.270.23
Mean NetworkBackbone=BEiT-3-base, Rel. Time=0.9×2026.0683.45.618.517.10.41—5.920.24
+ Temp. Sc.Backbone=BEiT-3-base, Rel. Time=0.9×2026.0683.45.618.517.10.36—1.210.23
StreamliningBackbone=BEiT-3-base, Rel. Time=2.1×2026.0683.45.668.214.40.37—1.380.23
4 samplesBackbone=BEiT-3-base, Rel. Time=4.0×2026.0683.15.551421.80.39—3.930.24
2 samplesBackbone=BEiT-3-base, Rel. Time=2.0×2026.0682.95.7810.719.90.4—4.480.24