A Transformer is trained for a factory robot that converts w…
Questions
A Trаnsfоrmer is trаined fоr а factоry robot that converts written instructions into actions. During testing, the model behaves almost identically on the following two commands:"Pick up the chemical container before opening the valve.""Open the valve before picking up the chemical container."The engineers confirm that: the vocabulary is adequate, the model was trained on many examples of both actions, and the self-attention layers are functioning normally.Which modification would most directly address this failure?