{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:Y4N2MKYRLD3Z24H5JP73YPBD3Q","short_pith_number":"pith:Y4N2MKYR","schema_version":"1.0","canonical_sha256":"c71ba62b1158f79d70fd4bffbc3c23dc10a2189e85ab2ed9b5042499457b42f9","source":{"kind":"arxiv","id":"2605.27932","version":1},"attestation_state":"computed","paper":{"title":"When Think-with-Image Meets Safety: What Determines Multimodal Jailbreak Robustness?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CR","cs.LG"],"primary_cat":"cs.CV","authors_text":"Binghang Lu, Bing Hu, Fang Wu, Neil Zhenqiang Gong, Xiaomin Li, Yuan Tian","submitted_at":"2026-05-27T04:04:42Z","abstract_excerpt":"Think-with-image reasoning is emerging as a new inference paradigm for large vision-language models, but its safety implications remain poorly understood. Existing systems already span multiple process designs, including direct response generation, text-only prior turn, visual-state manipulation, and explicit external image-tool invocation. In this paper, we ask which of these evaluated paradigms improves multimodal jailbreak robustness, and why. Across multiple vision-language models, explicit image-tool interaction yields the lowest attack success rates in our experiments, reducing jailbreak"},"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":"2605.27932","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-27T04:04:42Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CR","cs.LG"],"title_canon_sha256":"fdf5b01d8c84ed48688dbca81b423a78ae7f1d08d553adedc3c85aea000da195","abstract_canon_sha256":"4efc33224f361e533ab16ab783da028153096840b996d1e4ec559a88b3b0d6b5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:52.938349Z","signature_b64":"Tdk4V57J7Lf6jRf6pl9SKEEnz5DULD4ThhfQEwYQ+tmoSCO12HnXSx25cO7QwcIQsDGAYY8+OCdY9FJAApr9BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c71ba62b1158f79d70fd4bffbc3c23dc10a2189e85ab2ed9b5042499457b42f9","last_reissued_at":"2026-05-28T01:04:52.937926Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:52.937926Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When Think-with-Image Meets Safety: What Determines Multimodal Jailbreak Robustness?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CR","cs.LG"],"primary_cat":"cs.CV","authors_text":"Binghang Lu, Bing Hu, Fang Wu, Neil Zhenqiang Gong, Xiaomin Li, Yuan Tian","submitted_at":"2026-05-27T04:04:42Z","abstract_excerpt":"Think-with-image reasoning is emerging as a new inference paradigm for large vision-language models, but its safety implications remain poorly understood. Existing systems already span multiple process designs, including direct response generation, text-only prior turn, visual-state manipulation, and explicit external image-tool invocation. In this paper, we ask which of these evaluated paradigms improves multimodal jailbreak robustness, and why. Across multiple vision-language models, explicit image-tool interaction yields the lowest attack success rates in our experiments, reducing jailbreak"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.27932","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/2605.27932/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":"2605.27932","created_at":"2026-05-28T01:04:52.937991+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.27932v1","created_at":"2026-05-28T01:04:52.937991+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.27932","created_at":"2026-05-28T01:04:52.937991+00:00"},{"alias_kind":"pith_short_12","alias_value":"Y4N2MKYRLD3Z","created_at":"2026-05-28T01:04:52.937991+00:00"},{"alias_kind":"pith_short_16","alias_value":"Y4N2MKYRLD3Z24H5","created_at":"2026-05-28T01:04:52.937991+00:00"},{"alias_kind":"pith_short_8","alias_value":"Y4N2MKYR","created_at":"2026-05-28T01:04:52.937991+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/Y4N2MKYRLD3Z24H5JP73YPBD3Q","json":"https://pith.science/pith/Y4N2MKYRLD3Z24H5JP73YPBD3Q.json","graph_json":"https://pith.science/api/pith-number/Y4N2MKYRLD3Z24H5JP73YPBD3Q/graph.json","events_json":"https://pith.science/api/pith-number/Y4N2MKYRLD3Z24H5JP73YPBD3Q/events.json","paper":"https://pith.science/paper/Y4N2MKYR"},"agent_actions":{"view_html":"https://pith.science/pith/Y4N2MKYRLD3Z24H5JP73YPBD3Q","download_json":"https://pith.science/pith/Y4N2MKYRLD3Z24H5JP73YPBD3Q.json","view_paper":"https://pith.science/paper/Y4N2MKYR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.27932&json=true","fetch_graph":"https://pith.science/api/pith-number/Y4N2MKYRLD3Z24H5JP73YPBD3Q/graph.json","fetch_events":"https://pith.science/api/pith-number/Y4N2MKYRLD3Z24H5JP73YPBD3Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Y4N2MKYRLD3Z24H5JP73YPBD3Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Y4N2MKYRLD3Z24H5JP73YPBD3Q/action/storage_attestation","attest_author":"https://pith.science/pith/Y4N2MKYRLD3Z24H5JP73YPBD3Q/action/author_attestation","sign_citation":"https://pith.science/pith/Y4N2MKYRLD3Z24H5JP73YPBD3Q/action/citation_signature","submit_replication":"https://pith.science/pith/Y4N2MKYRLD3Z24H5JP73YPBD3Q/action/replication_record"}},"created_at":"2026-05-28T01:04:52.937991+00:00","updated_at":"2026-05-28T01:04:52.937991+00:00"}