Learning Coherent Representations: A Topological Approach to Interpretability

In three linesNovel topological approach to interpretability in deep neural networks. Authors introduce 'coherence', a geometric property where each neuron responds to contiguous regions of state space. They propose Coh, a differentiable objective function based on Fréchet variance, validated on MNIST and BERT embeddings.

Summary generated by Claude — human-verified