A zero-hidden-layer Topological Auditor prunes linear shortcuts, forcing networks to higher geometric capacity (N>=16) for fairer representations and cutting counterfactual gender vulnerability from 21.18% to 7.66%.
Title resolution pending
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
-
Fairness is Not Flat: Geometric Phase Transitions Against Shortcut Learning
A zero-hidden-layer Topological Auditor prunes linear shortcuts, forcing networks to higher geometric capacity (N>=16) for fairer representations and cutting counterfactual gender vulnerability from 21.18% to 7.66%.