Pruning the Paradox: How CLIP's Most Informative Heads Enhance Performance While Amplifying Bias

Documenti analoghi

• Quantifying and Enabling the Interpretability of CLIP-like Models
  di: Madasu, Avinash, et al.
  Pubblicazione: (2024)
• Cultural Awareness in Vision-Language Models: A Cross-Country Exploration
  di: Madasu, Avinash, et al.
  Pubblicazione: (2025)
• ICSVR: Investigating Compositional and Syntactic Understanding in Video Retrieval Models
  di: Madasu, Avinash, et al.
  Pubblicazione: (2023)
• SocialCounterfactuals: Probing and Mitigating Intersectional Social Biases in Vision-Language Models with Counterfactual Examples
  di: Howard, Phillip, et al.
  Pubblicazione: (2023)
• Training-Free Mitigation of Language Reasoning Degradation After Multimodal Instruction Tuning
  di: Ratzlaff, Neale, et al.
  Pubblicazione: (2024)