Estimating the Causal Effects of Natural Logic Features in Transformer-Based NLI Models

Open in new window