Bias Leaves a Gradient Trail: Label-Free Bias Identification via Gradient Probes on Concept Decompositions

Documenti analoghi

• From Neural Activations to Concepts: A Survey on Explaining Concepts in Neural Networks
  di: Lee, Jae Hee, et al.
  Pubblicazione: (2023)
• Read Between the Layers: Leveraging Multi-Layer Representations for Rehearsal-Free Continual Learning with Pre-Trained Models
  di: Ahrens, Kyra, et al.
  Pubblicazione: (2023)
• Gradient Descent as a Shrinkage Operator for Spectral Bias
  di: Lucey, Simon
  Pubblicazione: (2025)
• Inductive Gradient Adjustment For Spectral Bias In Implicit Neural Representations
  di: Shi, Kexuan, et al.
  Pubblicazione: (2024)
• Unlearning Concepts in Diffusion Model via Concept Domain Correction and Concept Preserving Gradient
  di: Wu, Yongliang, et al.
  Pubblicazione: (2024)