Transformer-based models can be affected by sequence length learning, leading to reliance on non-textual features for classification.
Transformer-based models can be affected by sequence length learning, leading to reliance on non-textual features over important textual information.