{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:EKNLBNEM7ERCHD3C6HIICEDZ67","short_pith_number":"pith:EKNLBNEM","schema_version":"1.0","canonical_sha256":"229ab0b48cf922238f62f1d0811079f7ecbb5102022f66889f44abf01b8c0710","source":{"kind":"arxiv","id":"2602.14399","version":2},"attestation_state":"computed","paper":{"title":"Multi-Turn Adaptive Prompting Attack on Large Vision-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Feng Liu, In Chong Choi, Jiacheng Zhang, Yiliao Song","submitted_at":"2026-02-16T02:15:58Z","abstract_excerpt":"Multi-turn jailbreak attacks have proven effective against text-only large language models (LLMs), where malicious content is gradually introduced to bypass safety alignment. However, effectively extending such attacks to large vision-language models (LVLMs) remains underexplored. In this paper, we find that naively incorporating visual inputs can make multi-turn jailbreaks easier to defend against; for example, overly malicious visual content will easily trigger the defense mechanism in safety-aligned LVLMs, resulting in more conservative responses. Based on this finding, we propose multi-tur"},"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":"2602.14399","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-02-16T02:15:58Z","cross_cats_sorted":[],"title_canon_sha256":"465ed56c9c0db114f0f1d90ffaef30a6db359e75f5e2f3667b1a7bb8fa7be968","abstract_canon_sha256":"7ef6088ed30701470d2c6776b9989813aba974a748d9e53fdce1735e82009205"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T01:05:04.874692Z","signature_b64":"2oln3bfsG5OQaeKnOcrQZTRoWCJo3Ay7yvilyItuMkyJnpS7qoX/QmeU/SQJyzuyOLvHTpZ+f/NLkhwBevKkDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"229ab0b48cf922238f62f1d0811079f7ecbb5102022f66889f44abf01b8c0710","last_reissued_at":"2026-05-29T01:05:04.873650Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T01:05:04.873650Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multi-Turn Adaptive Prompting Attack on Large Vision-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Feng Liu, In Chong Choi, Jiacheng Zhang, Yiliao Song","submitted_at":"2026-02-16T02:15:58Z","abstract_excerpt":"Multi-turn jailbreak attacks have proven effective against text-only large language models (LLMs), where malicious content is gradually introduced to bypass safety alignment. However, effectively extending such attacks to large vision-language models (LVLMs) remains underexplored. In this paper, we find that naively incorporating visual inputs can make multi-turn jailbreaks easier to defend against; for example, overly malicious visual content will easily trigger the defense mechanism in safety-aligned LVLMs, resulting in more conservative responses. Based on this finding, we propose multi-tur"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.14399","kind":"arxiv","version":2},"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/2602.14399/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":"2602.14399","created_at":"2026-05-29T01:05:04.873785+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.14399v2","created_at":"2026-05-29T01:05:04.873785+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.14399","created_at":"2026-05-29T01:05:04.873785+00:00"},{"alias_kind":"pith_short_12","alias_value":"EKNLBNEM7ERC","created_at":"2026-05-29T01:05:04.873785+00:00"},{"alias_kind":"pith_short_16","alias_value":"EKNLBNEM7ERCHD3C","created_at":"2026-05-29T01:05:04.873785+00:00"},{"alias_kind":"pith_short_8","alias_value":"EKNLBNEM","created_at":"2026-05-29T01:05:04.873785+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/EKNLBNEM7ERCHD3C6HIICEDZ67","json":"https://pith.science/pith/EKNLBNEM7ERCHD3C6HIICEDZ67.json","graph_json":"https://pith.science/api/pith-number/EKNLBNEM7ERCHD3C6HIICEDZ67/graph.json","events_json":"https://pith.science/api/pith-number/EKNLBNEM7ERCHD3C6HIICEDZ67/events.json","paper":"https://pith.science/paper/EKNLBNEM"},"agent_actions":{"view_html":"https://pith.science/pith/EKNLBNEM7ERCHD3C6HIICEDZ67","download_json":"https://pith.science/pith/EKNLBNEM7ERCHD3C6HIICEDZ67.json","view_paper":"https://pith.science/paper/EKNLBNEM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.14399&json=true","fetch_graph":"https://pith.science/api/pith-number/EKNLBNEM7ERCHD3C6HIICEDZ67/graph.json","fetch_events":"https://pith.science/api/pith-number/EKNLBNEM7ERCHD3C6HIICEDZ67/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EKNLBNEM7ERCHD3C6HIICEDZ67/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EKNLBNEM7ERCHD3C6HIICEDZ67/action/storage_attestation","attest_author":"https://pith.science/pith/EKNLBNEM7ERCHD3C6HIICEDZ67/action/author_attestation","sign_citation":"https://pith.science/pith/EKNLBNEM7ERCHD3C6HIICEDZ67/action/citation_signature","submit_replication":"https://pith.science/pith/EKNLBNEM7ERCHD3C6HIICEDZ67/action/replication_record"}},"created_at":"2026-05-29T01:05:04.873785+00:00","updated_at":"2026-05-29T01:05:04.873785+00:00"}