@misc{indiciae30e2f9a739ac,
  title = {What Do VLMs NOTICE? A Mechanistic Interpretability Pipeline for Gaussian-Noise-free Text-Image Corruption and Evaluation},
  author = {Michal Golovanevsky and William Rudman and Vedant Palit and Ritambhara Singh and Carsten Eickhoff},
  year = {2025},
  url = {https://arxiv.org/abs/2406.16320},
  note = {Source identifier: 2406.16320}
}