We present the TAPAS contribution to the Shared Task on Statement\nVerification and Evidence Finding with Tables (SemEval 2021 Task 9, Wang et al.\n(2021)). SEM TAB FACT Task A is a classification task of recognizing if a\nstatement is entailed, neutral or refuted by the content of a given table. We\nadopt the binary TAPAS model of Eisenschlos et al. (2020) to this task. We\nlearn two binary classification models: A first model to predict if a statement\nis neutral or non-neutral and a second one to predict if it is entailed or\nrefuted. As the shared task training set contains only entailed or refuted\nexamples, we generate artificial neutral examples to train the first model.\nBoth models are pre-trained using a MASKLM objective, intermediate\ncounter-factual and synthetic data (Eisenschlos et al., 2020) and TABFACT (Chen\net al., 2020), a large table entailment dataset. We find that the artificial\nneutral examples are somewhat effective at training the first model, achieving\n68.03 test F1 versus the 60.47 of a majority baseline. For the second stage, we\nfind that the pre-training on the intermediate data and TABFACT improves the\nresults over MASKLM pre-training (68.03 vs 57.01).\n