{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XMBHM4NRGL55LWDWCOJQ2BAUTR","short_pith_number":"pith:XMBHM4NR","schema_version":"1.0","canonical_sha256":"bb027671b132fbd5d87613930d04149c510d7762691d424b980f5aa4ac3a209b","source":{"kind":"arxiv","id":"2502.05772","version":1},"attestation_state":"computed","paper":{"title":"Effective Black-Box Multi-Faceted Attacks Breach Vision Large Language Model Guardrails","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Jun Zhu, Lanqing Hong, Lichao Wang, Xiao Yang, Yijun Yang","submitted_at":"2025-02-09T04:21:27Z","abstract_excerpt":"Vision Large Language Models (VLLMs) integrate visual data processing, expanding their real-world applications, but also increasing the risk of generating unsafe responses. In response, leading companies have implemented Multi-Layered safety defenses, including alignment training, safety system prompts, and content moderation. However, their effectiveness against sophisticated adversarial attacks remains largely unexplored. In this paper, we propose MultiFaceted Attack, a novel attack framework designed to systematically bypass Multi-Layered Defenses in VLLMs. It comprises three complementary "},"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":"2502.05772","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2025-02-09T04:21:27Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"de3d4fdf001e507239e982048f0ba0a4337e6a314cdcf18b9328f11de2934200","abstract_canon_sha256":"18a5c378488a76c7aab5d181a92f4228a8ff30b22a0545b0aee1f6187b7dae3e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:11:43.170552Z","signature_b64":"OnFQ1dOV38bWIkO4OqS7xImAeTrcnzDE0HaYH/xo0uUQ1MA4KV/nSHSsTVZjvLut5inLahZtV+Gdm+l6oiJYCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bb027671b132fbd5d87613930d04149c510d7762691d424b980f5aa4ac3a209b","last_reissued_at":"2026-07-05T10:11:43.170066Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:11:43.170066Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Effective Black-Box Multi-Faceted Attacks Breach Vision Large Language Model Guardrails","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Jun Zhu, Lanqing Hong, Lichao Wang, Xiao Yang, Yijun Yang","submitted_at":"2025-02-09T04:21:27Z","abstract_excerpt":"Vision Large Language Models (VLLMs) integrate visual data processing, expanding their real-world applications, but also increasing the risk of generating unsafe responses. In response, leading companies have implemented Multi-Layered safety defenses, including alignment training, safety system prompts, and content moderation. However, their effectiveness against sophisticated adversarial attacks remains largely unexplored. In this paper, we propose MultiFaceted Attack, a novel attack framework designed to systematically bypass Multi-Layered Defenses in VLLMs. It comprises three complementary "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.05772","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/2502.05772/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":"2502.05772","created_at":"2026-07-05T10:11:43.170125+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.05772v1","created_at":"2026-07-05T10:11:43.170125+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.05772","created_at":"2026-07-05T10:11:43.170125+00:00"},{"alias_kind":"pith_short_12","alias_value":"XMBHM4NRGL55","created_at":"2026-07-05T10:11:43.170125+00:00"},{"alias_kind":"pith_short_16","alias_value":"XMBHM4NRGL55LWDW","created_at":"2026-07-05T10:11:43.170125+00:00"},{"alias_kind":"pith_short_8","alias_value":"XMBHM4NR","created_at":"2026-07-05T10:11:43.170125+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XMBHM4NRGL55LWDWCOJQ2BAUTR","json":"https://pith.science/pith/XMBHM4NRGL55LWDWCOJQ2BAUTR.json","graph_json":"https://pith.science/api/pith-number/XMBHM4NRGL55LWDWCOJQ2BAUTR/graph.json","events_json":"https://pith.science/api/pith-number/XMBHM4NRGL55LWDWCOJQ2BAUTR/events.json","paper":"https://pith.science/paper/XMBHM4NR"},"agent_actions":{"view_html":"https://pith.science/pith/XMBHM4NRGL55LWDWCOJQ2BAUTR","download_json":"https://pith.science/pith/XMBHM4NRGL55LWDWCOJQ2BAUTR.json","view_paper":"https://pith.science/paper/XMBHM4NR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.05772&json=true","fetch_graph":"https://pith.science/api/pith-number/XMBHM4NRGL55LWDWCOJQ2BAUTR/graph.json","fetch_events":"https://pith.science/api/pith-number/XMBHM4NRGL55LWDWCOJQ2BAUTR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XMBHM4NRGL55LWDWCOJQ2BAUTR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XMBHM4NRGL55LWDWCOJQ2BAUTR/action/storage_attestation","attest_author":"https://pith.science/pith/XMBHM4NRGL55LWDWCOJQ2BAUTR/action/author_attestation","sign_citation":"https://pith.science/pith/XMBHM4NRGL55LWDWCOJQ2BAUTR/action/citation_signature","submit_replication":"https://pith.science/pith/XMBHM4NRGL55LWDWCOJQ2BAUTR/action/replication_record"}},"created_at":"2026-07-05T10:11:43.170125+00:00","updated_at":"2026-07-05T10:11:43.170125+00:00"}