representational alignment

Representational alignment refers to the degree to which different representations (e.g., learned embeddings) correspond with one another across different domains or tasks, facilitating transfer learning and collaboration between models.

6 papers