{
  "id": 4413,
  "url": "https://arxiv.org/abs/2605.13047v1",
  "title": "Revealing the Gap in Human and VLM Scene Perception through Counterfactual Semantic Saliency",
  "summary": "Evaluating whether large vision-language models (VLMs) align with human perception for high-level semantic scene comprehension remains a challenge. Traditional white-box interpretability methods are inapplicable to closed-source architectures and passive metrics fail to isolate causal features. We introduce Counterfactual Semantic Saliency (CSS). This black-box, model-agnostic framework quantifies the importance of objects by measuring the semantic shift induced by their causal ablation from a s",
  "authors": "Ziqi Wen, Parsa Madinei, Miguel P. Eckstein",
  "category": "research",
  "topics": "safety-alignment",
  "orgs": null,
  "regions": null,
  "published_at": "2026-05-13T06:11:47.000Z",
  "fetched_at": "2026-07-14T16:30:59.236Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/4413",
  "original_url": "https://arxiv.org/abs/2605.13047v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}