Interpretability Without Tradeoffs: Disentangling Polysemanticity At Equal Predictive Performance 事件

Interpretability Without Tradeoffs: Disentangling Polysemanticity At Equal Predictive Performance arXiv:2605.31304v1 Announce Type: cross Abstract: Deep neural networks (DNNs) are widely used, but interpreting what they actually learn remains difficult. A major obstacle is that individual neurons often encode multiple unrelated concepts, obscuring the decision process of the network. While prior work, such as sparse autoencoders, can separate these mixed signals into more meaningful, "monoseman