Bias Model Sample Clauses
Bias Model. As mentioned in Section 4.1.3, we train the bias model with spurious features. For MNLI and QQP, we adopt the hand-crafted word overlapping features proposed by [5], which includes: • Whether all the hypothesis words also belong to the premise. 7▇▇▇▇▇://▇▇▇▇▇▇▇.▇▇▇/ 8▇▇▇▇▇://▇▇▇▇▇▇.▇▇▇/huggingface/transformers MNLI HANS QQP PAWSqqp PAWSwiki FEVER Symm1 Symm2 ent 33.3% 50% dulp 36.9% 31.5% 44.2% supp 41.4% - - Train cont 33.3% neutral 33.3% 50% 0% Train non-dulp 63.1% 68.5% 55.8% Train refute not-info 17.2% 41.4% - - - -
Bias Model. As mentioned in Section 4.1.3, we train the bias model with spurious features. For MNLI and QQP, we adopt the hand-crafted word overlapping features proposed by [3], which includes: • Whether all the hypothesis words also belong to the premise. • Whether the hypothesis appears as a continuous subsequence in the premise. • The percentage of the hypothesis words wh = {wh, wh, · · · , wh h } that appear in the