Routing Probes Can Improve Without New Information: An Exact-Null Audit of Uncertainty Beyond Model Outputs
Routing signals of modern vision transformers -- expert gates, attention-residual weights and halting scores -- often improve probes that predict whether the model is correct, and the improvement is commonly read as evidence that routing carries information about errors beyond the model's outputs. We test this inferenc...
Wenhao Liang, Lin Yue, W. Zhang et al.
· 0 citations