[Pigalowda]

This reminds me a bit of the criticism of “black box logic” for ML models.

Is there something analogous to saliency maps for LLM?