PhD student, Frankfurt Institute for Advanced Studies
1 paper at NeurIPS 2025
In human vision, representational dimensions continually expand to enable abstraction, whereas artificial networks undergo late-stage collapse that may constrain their capacity for flexible generalization.
