{
  "url": "https://appliedfrontier.nl/en/research/signal-9ce097f6edce91e0/",
  "title": "Textual Response-Aware Guidance for Zero-Shot Adversarial Robustness of Vision-Language Models",
  "published": "2026-10-29",
  "excerpt": "Vision-language models (VLMs) exhibit strong zero-shot generalization but remain vulnerable to adversarial attacks. Existing adversarial fine-tuning methods often use pre-trained model guidance to preserve image--text",
  "content_sha256": "7d0fb77a8048e20ec36e6debfec25a2533168279f65fc2002f6fc03f369e4e30",
  "extracted_chars": 1398,
  "artifact_links": [
    "https://arxiv.org/abs/2610.40309"
  ],
  "arxiv_id": "2610.40309",
  "extraction_scope": "preprint_abstract_metadata",
  "full_text_available": false,
  "id": "signal-9ce097f6edce91e0",
  "org": "Research community",
  "topic": "Evaluation & safety",
  "source_kind": "preprint",
  "status": "unreviewed-source-report",
  "analysis_method": "deterministic_research_triage_v2",
  "novelty_verified": false,
  "proof_checked": false,
  "peer_review_verified": false,
  "matched_terms": [
    "Evaluation & safety"
  ],
  "checked_at": "2026-11-01T13:17:05+00:00",
  "first_seen": "2026-10-31T19:32:55+00:00",
  "source_hash": "8ab780eb2fc036d6ecb3045063a045cf7ff41f0171b0026466dd927d4c8fc227",
  "source_url": "https://arxiv.org/abs/2610.40309",
  "language": "en",
  "indexing": "noindex_until_editorially_enriched"
}