{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:INQWJVESGIEKHQLWEGT2QARBZW","short_pith_number":"pith:INQWJVES","schema_version":"1.0","canonical_sha256":"436164d4923208a3c17621a7a80221cd845c32615f3378845562528e3abfc2ab","source":{"kind":"arxiv","id":"2505.21967","version":1},"attestation_state":"computed","paper":{"title":"Seeing the Threat: Vulnerabilities in Vision-Language Models to Adversarial Attack","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Juan Ren, Mark Dras, Usman Naseem","submitted_at":"2025-05-28T04:43:39Z","abstract_excerpt":"Large Vision-Language Models (LVLMs) have shown remarkable capabilities across a wide range of multimodal tasks. However, their integration of visual inputs introduces expanded attack surfaces, thereby exposing them to novel security vulnerabilities. In this work, we conduct a systematic representational analysis to uncover why conventional adversarial attacks can circumvent the safety mechanisms embedded in LVLMs. We further propose a novel two stage evaluation framework for adversarial attacks on LVLMs. The first stage differentiates among instruction non compliance, outright refusal, and su"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2505.21967","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-28T04:43:39Z","cross_cats_sorted":[],"title_canon_sha256":"9c61eb858b321f3b005e22748b59b7208be0c063595a36a12a98cea9976488af","abstract_canon_sha256":"68135ccdb2f30e075574bbdf1f20483a66ea0df61dc80bf3dac9a7d3a7cc7979"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:11:07.028577Z","signature_b64":"hUi6phPY9KYoAY2Gk6WggH+IZ+WBibRuwk6LhKbhV5PaeU/PAFk2qjqvbMZnS2v1oTQmMjRegkiBSq3Uebr4Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"436164d4923208a3c17621a7a80221cd845c32615f3378845562528e3abfc2ab","last_reissued_at":"2026-07-05T11:11:07.028026Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:11:07.028026Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Seeing the Threat: Vulnerabilities in Vision-Language Models to Adversarial Attack","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Juan Ren, Mark Dras, Usman Naseem","submitted_at":"2025-05-28T04:43:39Z","abstract_excerpt":"Large Vision-Language Models (LVLMs) have shown remarkable capabilities across a wide range of multimodal tasks. However, their integration of visual inputs introduces expanded attack surfaces, thereby exposing them to novel security vulnerabilities. In this work, we conduct a systematic representational analysis to uncover why conventional adversarial attacks can circumvent the safety mechanisms embedded in LVLMs. We further propose a novel two stage evaluation framework for adversarial attacks on LVLMs. The first stage differentiates among instruction non compliance, outright refusal, and su"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.21967","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2505.21967/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2505.21967","created_at":"2026-07-05T11:11:07.028088+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.21967v1","created_at":"2026-07-05T11:11:07.028088+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.21967","created_at":"2026-07-05T11:11:07.028088+00:00"},{"alias_kind":"pith_short_12","alias_value":"INQWJVESGIEK","created_at":"2026-07-05T11:11:07.028088+00:00"},{"alias_kind":"pith_short_16","alias_value":"INQWJVESGIEKHQLW","created_at":"2026-07-05T11:11:07.028088+00:00"},{"alias_kind":"pith_short_8","alias_value":"INQWJVES","created_at":"2026-07-05T11:11:07.028088+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.10764","citing_title":"Break the Brake, Not the Wheel: Untargeted Jailbreak via Entropy Maximization","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10764","citing_title":"Break the Brake, Not the Wheel: Untargeted Jailbreak via Entropy Maximization","ref_index":28,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/INQWJVESGIEKHQLWEGT2QARBZW","json":"https://pith.science/pith/INQWJVESGIEKHQLWEGT2QARBZW.json","graph_json":"https://pith.science/api/pith-number/INQWJVESGIEKHQLWEGT2QARBZW/graph.json","events_json":"https://pith.science/api/pith-number/INQWJVESGIEKHQLWEGT2QARBZW/events.json","paper":"https://pith.science/paper/INQWJVES"},"agent_actions":{"view_html":"https://pith.science/pith/INQWJVESGIEKHQLWEGT2QARBZW","download_json":"https://pith.science/pith/INQWJVESGIEKHQLWEGT2QARBZW.json","view_paper":"https://pith.science/paper/INQWJVES","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.21967&json=true","fetch_graph":"https://pith.science/api/pith-number/INQWJVESGIEKHQLWEGT2QARBZW/graph.json","fetch_events":"https://pith.science/api/pith-number/INQWJVESGIEKHQLWEGT2QARBZW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/INQWJVESGIEKHQLWEGT2QARBZW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/INQWJVESGIEKHQLWEGT2QARBZW/action/storage_attestation","attest_author":"https://pith.science/pith/INQWJVESGIEKHQLWEGT2QARBZW/action/author_attestation","sign_citation":"https://pith.science/pith/INQWJVESGIEKHQLWEGT2QARBZW/action/citation_signature","submit_replication":"https://pith.science/pith/INQWJVESGIEKHQLWEGT2QARBZW/action/replication_record"}},"created_at":"2026-07-05T11:11:07.028088+00:00","updated_at":"2026-07-05T11:11:07.028088+00:00"}