List your company

Natural Language Understanding on GLUE (MNLI, MRPC, QNLI, QQP, SST2) (test)

88.08MNLI AccuracyCOIN
76.338479.386782.43585.4833Feb 9, 2026

Evaluation results

MethodLinks
COINPerturbation=CheckList, Backbone=Qwen2026.0288.08-0.3875.540.9274.382.7178.212.6292.67-0.1281.771.15
COINPerturbation=Clean, Backbone=Qwen2026.0287.71—76.46—77.08—80.83—92.54—82.93—
COINPerturbation=TextFolder, Backbone=Qwen2026.0287.580.1275.880.5873.333.7571.799.0491.54180.032.9
COINPerturbation=Avg, Backbone=Qwen2026.0287.480.2876.180.3575.841.5575.886.19920.6881.481.81
CoIPOPerturbation=TextFolder, Backbone=Qwen2026.0287.21-0.4672.626.2177.461.2984.040.0490182.271.62
COINPerturbation=DeepWordBug, Backbone=Qwen2026.0287.080.6276.83-0.3879.62-2.5473.876.9691.331.2181.751.18
COINPerturbation=StressTest, Backbone=Qwen2026.0286.960.7576.170.2974.792.2974.716.1291.920.6280.912.02
CoIPOPerturbation=Clean, Backbone=Qwen2026.0286.75—78.83—78.75—84.08—91—83.88—
CoIPOPerturbation=Avg, Backbone=Qwen2026.0286.520.2977.561.5978.77-0.0383.610.5990.790.2683.450.54
CoIPOPerturbation=DeepWordBug, Backbone=Qwen2026.0286.380.3778.83081.33-2.5882.08290.960.0483.92-0.03
CoIPOPerturbation=StressTest, Backbone=Qwen2026.0286.210.5478.120.7177.541.2183.540.5490.620.3883.210.68
CoIPOPerturbation=CheckList, Backbone=Qwen2026.0286.040.7179.38-0.5478.79-0.0484.29-0.2191.37-0.3783.97-0.09
BasePerturbation=Clean, Backbone=Qwen2026.0283.29—61.16—71.92—70.67—89.21—75.25—
BasePerturbation=StressTest, Backbone=Qwen2026.0282.750.5452.17969.122.7964.126.5482.56.7170.135.12
BasePerturbation=CheckList, Backbone=Qwen2026.0281.581.7164.83-3.6765.176.7574.21-3.5490.25-1.0475.210.04
SFTPerturbation=TextFolder, Backbone=Qwen2026.0280.83-0.6271.083.0467.924.0468.464.0886.674.2174.992.95
SFTPerturbation=CheckList, Backbone=Qwen2026.0280.62-0.4273.710.4271.50.4666.54691.42-0.5576.761.18
SFTPerturbation=Avg, Backbone=Qwen2026.0280.34-0.1772.881.5571.97-0.0270.023.1689.042.2976.851.36
BasePerturbation=Avg, Backbone=Qwen2026.0280.323.7258.053.8968.943.7268.27385.624.4972.243.76
SFTPerturbation=Clean, Backbone=Qwen2026.0280.21—74.12—71.96—72.54—90.87—77.94—
SFTPerturbation=StressTest, Backbone=Qwen2026.0280.21073.081.0471.710.2571.251.2986.424.4676.531.41
SFTPerturbation=DeepWordBug, Backbone=Qwen2026.0279.830.3772.421.7176.79-4.8371.291.2589.831.0478.03-0.09
BasePerturbation=DeepWordBug, Backbone=Qwen2026.0277.176.1353.757.4270.881.0470.83-0.1788.041.1772.133.12
BasePerturbation=TextFolder, Backbone=Qwen2026.0276.796.558.332.8367.624.2961.59.1778.0811.1268.476.78