Geometric features from per-layer MLP update trajectories fed to a sparse linear probe outperform maximum softmax probability for uncertainty quantification under selective abstention, with gains up to 21 AURC points.
arXiv preprint arXiv:2506.08572 , year=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Reading Calibrated Uncertainty from Language Model Trajectories
Geometric features from per-layer MLP update trajectories fed to a sparse linear probe outperform maximum softmax probability for uncertainty quantification under selective abstention, with gains up to 21 AURC points.